Datasets:

Modalities:
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
pandas
License:
Dataset Viewer
Auto-converted to Parquet
dataset
stringclasses
2 values
data_source
stringclasses
1 value
prompt
listlengths
2
2
ability
stringclasses
1 value
reward_model
dict
extra_info
dict
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What year saw the creation of the region where the county of Hertfordshire is located? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1994", "style": "rule" }
{ "answer": "1994", "evidence_list": [], "index": 1, "need_tools_kwargs": true, "question": "What year saw the creation of the region where the county of Hertfordshire is located?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1994" } }, "scrape": { "create_kwargs": { "ground_truth": "1994" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "When was the abolishment of the studio that distributed The Game? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1999", "style": "rule" }
{ "answer": "1999", "evidence_list": [], "index": 2, "need_tools_kwargs": true, "question": "When was the abolishment of the studio that distributed The Game?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1999" } }, "scrape": { "create_kwargs": { "ground_truth": "1999" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "Where did the Baldevins bryllup director die? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Copenhagen", "style": "rule" }
{ "answer": "Copenhagen", "evidence_list": [], "index": 7, "need_tools_kwargs": true, "question": "Where did the Baldevins bryllup director die?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Copenhagen" } }, "scrape": { "create_kwargs": { "ground_truth": "Copenhagen" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "Which major Russian city borders the body of water in which Saaremaa is located? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Saint Petersburg", "style": "rule" }
{ "answer": "Saint Petersburg", "evidence_list": [], "index": 9, "need_tools_kwargs": true, "question": "Which major Russian city borders the body of water in which Saaremaa is located?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Saint Petersburg" } }, "scrape": { "create_kwargs": { "ground_truth": "Saint Petersburg" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "When was the employer of John J. Collins established? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1822", "style": "rule" }
{ "answer": "1822", "evidence_list": [], "index": 10, "need_tools_kwargs": true, "question": "When was the employer of John J. Collins established?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1822" } }, "scrape": { "create_kwargs": { "ground_truth": "1822" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "In what year was the author of The Insider's Guide to the Colleges established? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1878", "style": "rule" }
{ "answer": "1878", "evidence_list": [], "index": 14, "need_tools_kwargs": true, "question": "In what year was the author of The Insider's Guide to the Colleges established?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1878" } }, "scrape": { "create_kwargs": { "ground_truth": "1878" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "Where did the director who provided the lyrics to A Time for Singing attend school? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Northwestern's School of Communication", "style": "rule" }
{ "answer": "Northwestern's School of Communication", "evidence_list": [], "index": 17, "need_tools_kwargs": true, "question": "Where did the director who provided the lyrics to A Time for Singing attend school?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Northwestern's School of Communication" } }, "scrape": { "create_kwargs": { "ground_truth": "Northwestern's School of Communication" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What languages are spoken, written or signed by the person the Schiff test is named after? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Italian", "style": "rule" }
{ "answer": "Italian", "evidence_list": [], "index": 32, "need_tools_kwargs": true, "question": "What languages are spoken, written or signed by the person the Schiff test is named after?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Italian" } }, "scrape": { "create_kwargs": { "ground_truth": "Italian" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What is the language for the person who translated Plato's work to Latin? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Italian", "style": "rule" }
{ "answer": "Italian", "evidence_list": [], "index": 36, "need_tools_kwargs": true, "question": "What is the language for the person who translated Plato's work to Latin?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Italian" } }, "scrape": { "create_kwargs": { "ground_truth": "Italian" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "Who was the president of the body that petitioned the king of England to recognize the rights of colonists when the declaration of independence was adopted? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "President John Hancock", "style": "rule" }
{ "answer": "President John Hancock", "evidence_list": [], "index": 38, "need_tools_kwargs": true, "question": "Who was the president of the body that petitioned the king of England to recognize the rights of colonists when the declaration of independence was adopted?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "President John Hancock" } }, "scrape": { "create_kwargs": { "ground_truth": "President John Hancock" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "In what year did France host the finals of the competition featuring Samoa for going on 30 years? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "2007", "style": "rule" }
{ "answer": "2007", "evidence_list": [], "index": 41, "need_tools_kwargs": true, "question": "In what year did France host the finals of the competition featuring Samoa for going on 30 years?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "2007" } }, "scrape": { "create_kwargs": { "ground_truth": "2007" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "In which year did the publisher of Roald Dahl's Guide to Railway Safety cease to exist? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "2001", "style": "rule" }
{ "answer": "2001", "evidence_list": [], "index": 42, "need_tools_kwargs": true, "question": "In which year did the publisher of Roald Dahl's Guide to Railway Safety cease to exist?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "2001" } }, "scrape": { "create_kwargs": { "ground_truth": "2001" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "When was the inception of the company that released Poptropica? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1998", "style": "rule" }
{ "answer": "1998", "evidence_list": [], "index": 43, "need_tools_kwargs": true, "question": "When was the inception of the company that released Poptropica?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1998" } }, "scrape": { "create_kwargs": { "ground_truth": "1998" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What type of community is the municipality where Norbert Pfretzschner died, an instance of? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "comune", "style": "rule" }
{ "answer": "comune", "evidence_list": [], "index": 49, "need_tools_kwargs": true, "question": "What type of community is the municipality where Norbert Pfretzschner died, an instance of?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "comune" } }, "scrape": { "create_kwargs": { "ground_truth": "comune" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What conflict involved the most famous Confederacy army? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "American Civil War", "style": "rule" }
{ "answer": "American Civil War", "evidence_list": [], "index": 53, "need_tools_kwargs": true, "question": "What conflict involved the most famous Confederacy army?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "American Civil War" } }, "scrape": { "create_kwargs": { "ground_truth": "American Civil War" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What is the name of the airport in the city where the torch event was held in Vietnam? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Tân Sơn Nhất International Airport", "style": "rule" }
{ "answer": "Tân Sơn Nhất International Airport", "evidence_list": [], "index": 54, "need_tools_kwargs": true, "question": "What is the name of the airport in the city where the torch event was held in Vietnam?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Tân Sơn Nhất International Airport" } }, "scrape": { "create_kwargs": { "ground_truth": "Tân Sơn Nhất International Airport" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What year was the end of the manufacturer of the White armored car? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1980", "style": "rule" }
{ "answer": "1980", "evidence_list": [], "index": 57, "need_tools_kwargs": true, "question": "What year was the end of the manufacturer of the White armored car?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1980" } }, "scrape": { "create_kwargs": { "ground_truth": "1980" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What university is Anthony Marten's college part of? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "college of the University of Cambridge", "style": "rule" }
{ "answer": "college of the University of Cambridge", "evidence_list": [], "index": 59, "need_tools_kwargs": true, "question": "What university is Anthony Marten's college part of?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "college of the University of Cambridge" } }, "scrape": { "create_kwargs": { "ground_truth": "college of the University of Cambridge" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What university is the college attended by Roland Moyle part of? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "college of the University of Cambridge", "style": "rule" }
{ "answer": "college of the University of Cambridge", "evidence_list": [], "index": 61, "need_tools_kwargs": true, "question": "What university is the college attended by Roland Moyle part of?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "college of the University of Cambridge" } }, "scrape": { "create_kwargs": { "ground_truth": "college of the University of Cambridge" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What language is used by the director of Thenum Vayambum? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Hindi", "style": "rule" }
{ "answer": "Hindi", "evidence_list": [], "index": 72, "need_tools_kwargs": true, "question": "What language is used by the director of Thenum Vayambum?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Hindi" } }, "scrape": { "create_kwargs": { "ground_truth": "Hindi" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What is the family of the current leader of Saudi Arabia? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Al Saud", "style": "rule" }
{ "answer": "Al Saud", "evidence_list": [], "index": 76, "need_tools_kwargs": true, "question": "What is the family of the current leader of Saudi Arabia?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Al Saud" } }, "scrape": { "create_kwargs": { "ground_truth": "Al Saud" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What was the language of the designer of Sacro Cuore di Cristo Re? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Italian", "style": "rule" }
{ "answer": "Italian", "evidence_list": [], "index": 82, "need_tools_kwargs": true, "question": "What was the language of the designer of Sacro Cuore di Cristo Re?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Italian" } }, "scrape": { "create_kwargs": { "ground_truth": "Italian" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "Who was President of the U.S. following the event which was preceded by Joseph Schumpeter and Norman Angell's most prolific writing ? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Thomas Woodrow Wilson", "style": "rule" }
{ "answer": "Thomas Woodrow Wilson", "evidence_list": [], "index": 84, "need_tools_kwargs": true, "question": "Who was President of the U.S. following the event which was preceded by Joseph Schumpeter and Norman Angell's most prolific writing ?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Thomas Woodrow Wilson" } }, "scrape": { "create_kwargs": { "ground_truth": "Thomas Woodrow Wilson" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "When did Tito first visit the person who laid claim to Eritrea and Italian Somaliland in a letter to Franklin D. Roosevelt? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1954", "style": "rule" }
{ "answer": "1954", "evidence_list": [], "index": 85, "need_tools_kwargs": true, "question": "When did Tito first visit the person who laid claim to Eritrea and Italian Somaliland in a letter to Franklin D. Roosevelt?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1954" } }, "scrape": { "create_kwargs": { "ground_truth": "1954" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "Who is the child of Liszt's mistress? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Cosima Wagner", "style": "rule" }
{ "answer": "Cosima Wagner", "evidence_list": [], "index": 89, "need_tools_kwargs": true, "question": "Who is the child of Liszt's mistress?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Cosima Wagner" } }, "scrape": { "create_kwargs": { "ground_truth": "Cosima Wagner" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "When was the record label for Lovely Creatures created? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "2003", "style": "rule" }
{ "answer": "2003", "evidence_list": [], "index": 91, "need_tools_kwargs": true, "question": "When was the record label for Lovely Creatures created?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "2003" } }, "scrape": { "create_kwargs": { "ground_truth": "2003" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "When did the institute that gives out the Elliott Cresson Medal award open? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1824", "style": "rule" }
{ "answer": "1824", "evidence_list": [], "index": 93, "need_tools_kwargs": true, "question": "When did the institute that gives out the Elliott Cresson Medal award open?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1824" } }, "scrape": { "create_kwargs": { "ground_truth": "1824" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What year did the company that built K200 KIFV dissolve? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1999", "style": "rule" }
{ "answer": "1999", "evidence_list": [], "index": 94, "need_tools_kwargs": true, "question": "What year did the company that built K200 KIFV dissolve?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1999" } }, "scrape": { "create_kwargs": { "ground_truth": "1999" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What family does the mother of Louise of Great Britain belong to? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "House of Hohenzollern", "style": "rule" }
{ "answer": "House of Hohenzollern", "evidence_list": [], "index": 103, "need_tools_kwargs": true, "question": "What family does the mother of Louise of Great Britain belong to?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "House of Hohenzollern" } }, "scrape": { "create_kwargs": { "ground_truth": "House of Hohenzollern" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "Which family is the mother of Anne, Princess Royal and Princess of Orange's a member of? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "House of Hohenzollern", "style": "rule" }
{ "answer": "House of Hohenzollern", "evidence_list": [], "index": 110, "need_tools_kwargs": true, "question": "Which family is the mother of Anne, Princess Royal and Princess of Orange's a member of?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "House of Hohenzollern" } }, "scrape": { "create_kwargs": { "ground_truth": "House of Hohenzollern" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "Who was the first president from the party following the Anti-Administration Party? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Jefferson", "style": "rule" }
{ "answer": "Jefferson", "evidence_list": [], "index": 111, "need_tools_kwargs": true, "question": "Who was the first president from the party following the Anti-Administration Party?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Jefferson" } }, "scrape": { "create_kwargs": { "ground_truth": "Jefferson" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "During what years did Frédéric take lessons at the academy where Chopin's father once had a teaching position? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1823 to 1826", "style": "rule" }
{ "answer": "1823 to 1826", "evidence_list": [], "index": 113, "need_tools_kwargs": true, "question": "During what years did Frédéric take lessons at the academy where Chopin's father once had a teaching position?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1823 to 1826" } }, "scrape": { "create_kwargs": { "ground_truth": "1823 to 1826" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "In what year did the agency that maintains the Presidential Library system become independent? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1985", "style": "rule" }
{ "answer": "1985", "evidence_list": [], "index": 115, "need_tools_kwargs": true, "question": "In what year did the agency that maintains the Presidential Library system become independent?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1985" } }, "scrape": { "create_kwargs": { "ground_truth": "1985" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "In what year did the South Korean car manufacturer that purchased the factor in 1995 dissolve? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1999", "style": "rule" }
{ "answer": "1999", "evidence_list": [], "index": 116, "need_tools_kwargs": true, "question": "In what year did the South Korean car manufacturer that purchased the factor in 1995 dissolve?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1999" } }, "scrape": { "create_kwargs": { "ground_truth": "1999" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What was a notable work of the Chet Holifield Federal Building designer? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Transamerica Pyramid", "style": "rule" }
{ "answer": "Transamerica Pyramid", "evidence_list": [], "index": 117, "need_tools_kwargs": true, "question": "What was a notable work of the Chet Holifield Federal Building designer?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Transamerica Pyramid" } }, "scrape": { "create_kwargs": { "ground_truth": "Transamerica Pyramid" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "What establishment today contains what was known as the place Chopin's father got a teaching position during that time? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "Warsaw University", "style": "rule" }
{ "answer": "Warsaw University", "evidence_list": [], "index": 118, "need_tools_kwargs": true, "question": "What establishment today contains what was known as the place Chopin's father got a teaching position during that time?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "Warsaw University" } }, "scrape": { "create_kwargs": { "ground_truth": "Warsaw University" } } } }
musique_train
musique_train
[ { "content": "\nIn this environment you have access to a set of tools you can use to answer the user's question.\n\nYou only have access to the tools provided below. You can only use one tool per message, and will receive the result of that tool in the user's next response. You use tools step-by-step to accomplish a given task, with each tool-use informed by the result of the previous tool-use. Today is: 2025-07-16\n\n# Tool-Use Formatting Instructions\n\nTool-use is formatted using XML-style tags. The tool-use is enclosed in <use_mcp_tool></use_mcp_tool> and each parameter is similarly enclosed within its own set of tags.\n\nThe Model Context Protocol (MCP) connects to servers that provide additional tools and resources to extend your capabilities. You can use the server's tools via the `use_mcp_tool`.\n\nDescription:\nRequest to use a tool provided by a MCP server. Each MCP server can provide multiple tools with different capabilities. Tools have defined input schemas that specify required and optional parameters.\n\nParameters:\n- server_name: (required) The name of the MCP server providing the tool\n- tool_name: (required) The name of the tool to execute\n- arguments: (required) A JSON object containing the tool's input parameters, following the tool's input schema, quotes within string must be properly escaped, ensure it's valid JSON\n \nUsage:\n<use_mcp_tool>\n<server_name>server name here</server_name>\n<tool_name>tool name here</tool_name>\n<arguments>\n{\n \"param1\": \"value1\",\n \"param2\": \"value2 \"escaped string\"\"\n}\n</arguments>\n</use_mcp_tool>\n\nImportant Notes:\n- Tool-use must be placed **at the end** of your response, **top-level**, and not nested within other tags.\n- Always adhere to this format for the tool use to ensure proper parsing and execution.\n\nString and scalar parameters should be specified as is, while lists and objects should use JSON format. Note that spaces for string values are not stripped. The output is not expected to be valid XML and is parsed with regular expressions.\n\n\n\nHere are the functions available in JSONSchema format:\n\n## Server name: browsing_agent\n### Tool name: search_and_browse\nDescription: This tool is an agent that performs the subtask of searching and browsing the web for specific missing information and generating the desired answer. The subtask should be clearly defined, include relevant background, and focus on factual gaps. It does not perform vague or speculative subtasks. \nArgs: \n\tsubtask: the subtask to be performed. \nReturns: \n\tthe result of the subtask.\nInput JSON schema: {'properties': {'subtask': {'title': 'Subtask', 'type': 'string'}}, 'required': ['subtask'], 'title': 'search_and_browseArguments', 'type': 'object'}\n\n\n\n# General Objective\n\nYou accomplish a given task iteratively, breaking it down into clear steps and working through them methodically.\n\n## Task Strategy\n\n1. Analyze the user's request and set clear, achievable sub-goals. Prioritize these sub-goals in a logical order.\n2. Start with a concise, numbered, step-by-step plan outlining how you will solve the task before taking any action.\n3. Work through these sub-goals sequentially. After each step, adjust your plan as needed.\n4. Use tools strategically to accomplish each sub-goal.\n5. Revise earlier steps if new information emerges.\n\n## Tool-Use Guidelines\n\n1. Each step must involve a single tool call, unless the task is already solved.\n2. Before each tool call:\n - Summarize what is known.\n - Identify what is missing.\n - Choose the most relevant tool.\n - Verify all required parameters.\n3. All tool queries must include full context.\n4. Avoid vague queries. Each call should retrieve actionable information.\n5. Extract and summarize partial information if a tool result is incomplete.\n\n## Tool-Use Communication Rules\n\n1. Do not include tool results in your response.\n2. Do not present the final answer until the entire task is complete.\n3. Do not mention tool names.\n4. Do not engage in unnecessary back-and-forth.\n5. Do not use non-existent tools.\n6. Respond in the same language as the user's message.\n7. If the task does not require tool use, answer directly.\n\n\n# Agent Specific Objective\n\nYou are a task-solving agent that uses tools step-by-step to answer the user's question. Your goal is to provide complete, accurate and well-reasoned answers using additional tools.\n\n\n", "role": "system" }, { "content": "When did the sports team that Junior Félix played for win their last World Series? You should follow the format instruction in the requestion strictly and wrap the final answer in \\boxed{}.", "role": "user" } ]
qa
{ "ground_truth": "1993", "style": "rule" }
{ "answer": "1993", "evidence_list": [], "index": 120, "need_tools_kwargs": true, "question": "When did the sports team that Junior Félix played for win their last World Series?", "question_type": "inference_query", "split": "test", "tools_kwargs": { "google_search": { "create_kwargs": { "ground_truth": "1993" } }, "scrape": { "create_kwargs": { "ground_truth": "1993" } } } }
End of preview. Expand in Data Studio

MATPO: Multi-Agent Tool-Integrated Policy Optimization

Train Multiple Agent Roles Within a Single LLM via Reinforcement Learning.

Models Data Paper Github

GAIA Results
GAIA Results
FRAMES Results
FRAMES Results
WebWalkerQA Results
WebWalkerQA Results

MATPO Framework

MATPO allows planner and worker agents to coexist within a single LLM and be trained via RL, achieving an 18.38% relative improvement over single-agent baselines on GAIA-text, FRAMES, and WebWalker-QA.

News & Updates

  • [2025-Oct-08] MATPO-Qwen3-14B checkpoints and rollouts released
  • [2025-Oct-08] Code and training scripts released
  • [2025-Oct-06] Arxiv Paper released

Overview

MATPO (Multi-Agent Tool-Integrated Policy Optimization) is a novel reinforcement learning framework that enables training multiple specialized agent roles (planner and worker agents) within a single large language model.

The Problem

Current single-agent approaches for multi-turn tool-integrated planning face critical limitations:

  • Context Length Bottleneck: Tool responses (e.g., web scraping) consume excessive tokens, making long-range planning prohibitive
  • Noisy Tool Responses: Raw tool responses interfere with the model's attention and planning capabilities

Our Solution

MATPO introduces a multi-agent-in-one-model architecture where:

  • A planner-agent orchestrates high-level planning and delegates subtasks
  • Worker-agents handle specific browsing and search tasks with isolated contexts
  • Both roles are trained within a single LLM using role-specific prompts via reinforcement learning

Key Features

  • Multi-Agent-in-One-Model: Train planner and worker agents within a single LLM using role-specific system prompts
  • Principled Credit Assignment: Extends GRPO with theoretically grounded reward distribution across planner and worker rollouts
  • Easy Integration: Built on top of veRL, compatible with existing RL training frameworks
  • Robust Training: More stable learning curves compared to single-agent approaches, especially with noisy tool responses
  • Infrastructure Efficient: No need for deployment of separate models or additional rollout engines

MATPO Architecture

MATPO employs a hierarchical multi-agent framework where a single LLM serves multiple roles:

User Query → Planner Agent → Subtask 1 → Worker Agent → Result 1
                           → Subtask 2 → Worker Agent → Result 2
                           → ...
                           → Final Answer

Single-agent GRPO Framework MATPO Framework

Comparison between the rollout trajectories between the single-agent GRPO (top) and the multi-agent MATPO (bottom).

Multi-Agent Rollout Process

  1. Planner Agent:

    • Receives user query with planner-specific system prompt
    • Generates high-level plan and decomposes it into subtasks
    • Delegates subtasks to worker agents
    • Synthesizes worker responses into final answer
  2. Worker Agent:

    • Receives subtask with worker-specific system prompt
    • Performs multi-turn tool-integrated planning (search, scrape, analyze)
    • Returns summarized result to planner
    • Maintains isolated context to prevent token overflow
  3. Credit Assignment:

    • Final answer accuracy determines the reward
    • Reward is normalized across all planner-worker rollout groups
    • Gradient flows to both planner actions and worker actions proportionally

MATPO Framework

Visualization of MATPO implementation.

Quick Start

Prerequisites:

  • Python 3.10 or higher
  • CUDA 12.4+ (for GPU support)
  • 16 x (8 x 80G-A800) GPUs (for training with Qwen3-14B-base)

Clone the repository.

git clone https://github.com/mzf666/MATPO.git
cd MATPO

For prerequisites installation (CUDA, cuDNN, Apex), we recommend following the verl prerequisites guide which provides detailed instructions for:

  • CUDA: Version >= 12.4
  • cuDNN: Version >= 9.8.0
  • Apex

Setup environment and install dependencies.

conda create -n matpo python==3.10 -y
conda activate matpo
bash examples/sglang_multiturn/install.sh

Setup Node.js for Serper API support.

MCP (Model Context Protocol) requires Node.js to run MCP servers. Node.js version 18+ is recommended for optimal compatibility with MCP tools.

target_path=YOUR_TARGET_PATH

# Download Node.js binary (example for Linux x64)
wget https://nodejs.org/dist/v24.2.0/node-v24.2.0-linux-x64.tar.xz

# Extract to your target path
tar -xf node-v24.2.0-linux-x64.tar.xz -C $target_path

# Add to PATH
export NODEJS_HOME=$target_path/node-v24.2.0-linux-x64
export PATH=$NODEJS_HOME/bin:$PATH
export NODE_SHARED=$target_path/node-shared/node_modules
export PATH=$NODE_SHARED/.bin:$PATH

# Verify installation
node --version
npm --version

# Install serper mcp server
mkdir -p $target_path/node-shared
cd $target_path/node-shared
npm init -y
npm install serper-search-scrape-mcp-server

Configure the Node.js paths and HTTP / HTTPS proxies (if necessary) in the examples/sglang_multiturn/launch.sh script properly.

Download the training and testing datasets to the data directory. The prerpocessed datasets can be downloaded here.

Train a Qwen3-14B-base model with MATPO on the MuSiQue dataset and evaluate on the GAIA-text datasets:

# tested on 16 x (8 x 80G-A800) nodes

export SERPER_API_KEY="YOUR_SERPER_API_KEY" && \
export OPENAI_API_KEY="YOUR_OPENAI_API_KEY" && \
export WANDB_API_KEY="YOUR_WANDB_API_KEY" && \
export SINGLENODE=true && \
export RAY_DEBUG=legacy && \
export HYDRA_FULL_ERROR=1 && \
source YOUR_CONDA_PATH activate matpo && \
cd YOUR_PROJECT_PATH && \
bash examples/sglang_multiturn/launch.sh \
    examples/sglang_multiturn/qwen3-14b_musique_MATPO.sh

Experiments and Results

Main Results

MATPO consistently outperforms single-agent GRPO baselines across all benchmarks:

Method GAIA-text WebWalkerQA FRAMES Relative Average Improvement
Single-Agent GRPO 32.16% 30.14% 56.22% -
MATPO (Ours) 42.60% 33.00% 63.64% +18.38%

Training Configuration

  • Base Model: Qwen3-14B-base
  • Training Dataset: Filtered MuSiQue dataset.
  • Training Steps: 180 steps
  • Rollouts per Query: 8 (for group normalization)
  • Reward Function: 0.9 × accuracy + 0.1 × tool_format_reward

Model Checkpoints and Rollouts

We release the trained Qwen3-14B-base model checkpoints at the 180th training step of both single-agent GRPO and MATPO.

The associated model rollouts across various training steps can be found here.

Key Findings

  • More Stable Training: MATPO exhibits more stable learning curves and avoids catastrophic performance drops observed in single-agent training

  • Robustness to Noise: Multi-agent decomposition effectively isolates noisy tool responses, preventing them from interfering with high-level planning

  • Better Credit Assignment: Principled reward distribution across planner and worker rollouts leads to more effective learning

Practical Implementation Tips

Based on our experiments, we recommend:

  • Final Summary: Final summaries from worker agents are critical for clean planner-worker interfaces
  • Query Recap: Recapping original user query in worker prompt significantly improves performance
  • URL Blocking: Remember to blocking HuggingFace search results to avoid data leakage

Citation

If you find MATPO helpful in your research, please consider citing our paper:

@misc{mo2025multiagenttoolintegratedpolicyoptimization,
      title={Multi-Agent Tool-Integrated Policy Optimization}, 
      author={Zhanfeng Mo and Xingxuan Li and Yuntao Chen and Lidong Bing},
      year={2025},
      eprint={2510.04678},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2510.04678}, 
}

Acknowledgments

We would like to thank:

  • VolcEngine for developing and open-sourcing veRL, the RL training framework that powers MATPO
  • Alibaba Cloud for the Qwen3 model series
  • Google for the Serper API that enables web search capabilities
  • The authors of GAIA, WebWalkerQA, FRAMES, and MuSiQue datasets
  • The open-source community for valuable feedback and contributions

FAQ

Q: What's the difference between MATPO and traditional multi-agent systems?

MATPO uses a single LLM to play multiple agent roles via different system prompts, rather than deploying separate models. This offers:

  • Lower infrastructure complexity
  • Better parameter efficiency
  • Easier deployment and maintenance
  • Compatible with existing RL frameworks
Q: Can I use MATPO with models other than Qwen3?

Yes! MATPO is model-agnostic. You can use any decoder-only LLM that supports tool calling and multi-turn conversations. We've tested with Qwen3-14B-base, but models like Llama 3, Mistral, or other reasoning-capable LLMs should work.

Q: How many GPUs do I need for training?

For Qwen3-14B-base, we recommend:

  • Training: 8x A100/A800 GPUs (80GB)
  • Inference: 1-2x A100/A800 GPUs (40GB/80GB)
Q: How does MATPO handle credit assignment?

MATPO extends GRPO with principled credit assignment:

  1. The planner's final answer determines the accuracy reward
  2. This reward is normalized across all rollouts in a group
  3. Gradients flow proportionally to both planner and worker actions
  4. Worker agents receive the same advantage value as their parent planner rollout

See our paper for more details.

Q: Can I use MATPO for tasks other than web search?

Absolutely! While our paper focuses on web search, MATPO's framework is general. You can extend it to:

  • Code generation with execution feedback
  • Scientific reasoning with calculator tools
  • Data analysis with pandas/SQL tools
  • Any multi-turn task with verifiable rewards
Q: How stable is MATPO training compared to single-agent RL?

MATPO is significantly more stable. Our experiments show:

  • Single-agent GRPO often suffers catastrophic drops after step 120
  • MATPO maintains steady improvement throughout training
  • Multi-agent structure isolates noisy tool responses, preventing interference

See Figure 4 in our paper for training curves.

Q: Do I need to block HuggingFace URLs during training?

For research integrity, yes - especially if your evaluation benchmarks are hosted on HuggingFace. This prevents models from "cheating" by finding ground-truth answers online.

For production systems with no data leakage concerns, this is optional.


Star ⭐ this repository if you find it helpful!

Downloads last month
610