Hugging Face's logo

Mathoctopus
/
Mistral_Parallel_7B

Text Generation
Transformers
PyTorch
mistral
text-generation-inference
Model card Files
xet
Community
1
Nuo97 commited on
Commit
179a9a0
·
verified ·
1 Parent(s): e0b1b4f

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +157 -0
README.md CHANGED
@@ -1,3 +1,160 @@
1
  ---
2
  license: apache-2.0
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: apache-2.0
3
  ---
4
+ ---
5
+ license: apache-2.0
6
+ datasets:
7
+ - Mathoctopus/GSM8KInstruct_Parallel
8
+ language:
9
+ - en
10
+ - es
11
+ - zh
12
+ - de
13
+ - ru
14
+ - th
15
+ - sw
16
+ - ja
17
+ - fr
18
+ - bn
19
+ ---
20
+
21
+ # 🐙 Breaking Language Barriers in Multilingual Mathematical Reasoning: Insights and Observations
22
+
23
+ Project Page: [https://mathoctopus.github.io/](https://mathoctopus.github.io/)
24
+
25
+ Paper: [https://arxiv.org/abs/2310.20246.pdf](https://arxiv.org/abs/2310.20246.pdf)
26
+
27
+ Code: [https://github.com/microsoft/MathOctopus](https://github.com/microsoft/MathOctopus)
28
+
29
+ ### Introduction
30
+
31
+ We introduce 🐙 MathOctopus, a series of open-source large language models (LLMs) specifically tailored for multilingual math problem-solving. The MathOctopus models are trained on 🤗 MGSM8KInstruct Dataset, encompassing ten distinct languages.
32
+ MathOctopus notably outperforms conventional open-source LLMs and exhibits superiority over ChatGPT in few-shot scenarios.
33
+
34
+ ### Datasets
35
+
36
+ #### **MGSM8KInstruct**
37
+
38
+ | Training Dataset | En | Sw | Zh | Bn | De | Es | Fr | Ja | Ru | Th | Overall |
39
+ |:----------------------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|
40
+ | MGSM8KInstruct | 7473 | 7472 | 7466 | 6539 | 7466 | 7470 | 7469 | 7471 | 7361 | 7473 | **73.6K** |
41
+
42
+
43
+ #### **MSVAMP**
44
+
45
+ | Test Dataset | En | Sw | Zh | Bn | De | Es | Fr | Ja | Ru | Th | Overall |
46
+ |:----------------------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|
47
+ | MSVAMP | 1000 | 1000 | 1000 | 1000 | 1000 | 1000 | 1000 | 1000 | 1000 | 1000 | **10K** |
48
+
49
+ #### Usage
50
+
51
+ Our dataset and models are all available at Huggingface.
52
+
53
+ 🤗 [MGSM8KInstruct_Parallel Dataset](https://huggingface.co/datasets/Mathoctopus/GSM8KInstruct_Parallel)
54
+
55
+ 🤗 [MGSM8KInstruct_Cross Dataset](https://huggingface.co/datasets/Mathoctopus/MGSM8KInstruct_Cross)
56
+
57
+ 🤗 [MSVAMP Dataset](https://huggingface.co/datasets/Mathoctopus/MSVAMP)
58
+
59
+
60
+ ## Models
61
+
62
+ | Base Model: LLama | Parallel-Training | Cross-Training |
63
+ |----|---------------------------------------------------------------|---------------------------------------------------------------------------|
64
+ | 7B-LLaMA 2 | 🐙 [MathOctopus-Parallel-7B](https://huggingface.co/Mathoctopus/Parallel_7B) | 🐙 [MathOctopus-Cross-7B](https://huggingface.co/Mathoctopus/Cross_7B) |
65
+ || 🐙[MathOctopus-Parallel-xRFT-7B](https://huggingface.co/Mathoctopus/Parallel_xRFT_7B)|🐙[MathOctopus-Cross-xRFT-7B](https://huggingface.co/Mathoctopus/Cross_xRFT_7B)|
66
+ | 13B-LLaMA 2 | 🐙 [MathOctopus-Parallel-13B](https://huggingface.co/Mathoctopus/Parallel_13B) | 🐙 [MathOctopus-Cross-13B](https://huggingface.co/Mathoctopus/Cross_13B) |
67
+ || 🐙[MathOctopus-Parallel-xRFT-13B](https://huggingface.co/Mathoctopus/Parallel_xRFT_13B)|🐙[MathOctopus-Cross-xRFT-13B](https://huggingface.co/Mathoctopus/Cross_xRFT_13B/tree/main)|
68
+ | 33B-LLaMA 1 | 🐙 [MathOctopus-Parallel-33B](https://huggingface.co/Mathoctopus/Parallel_33B) | 🐙 [MathOctopus-Cross-33B](https://huggingface.co/Mathoctopus/Cross_33B/tree/main) |
69
+ | 70B-LLaMA 2 | Coming soon! | Coming Soon! |
70
+
71
+ *-Parallel refers to our model trained with the parallel-training strategy.
72
+
73
+ *-Cross refers to our model trained with cross-training strategy.
74
+
75
+ *-xRFT means we train the model with multilingual rejection sampling.
76
+
77
+ ### **Overall Results on MGSM**
78
+
79
+ | 7B Model | En | Sw | Zh | Bn | De | Es | Fr | Ja | Ru | Th | Overall |
80
+ |:--------------------------------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|
81
+ | MathOctopus<sup>C</sup> | 52.0 | 23.6 | 31.6 | 18.8 | 38.0 | 39.2 | 36.4 | 27.2 | 33.6 | 21.6 | 32.2 |
82
+ | **xRFT**-MathOctopus<sup>C</sup>| 51.2 | 24.0 | 33.2 | 18.8 | 36.0 | 41.2 | 37.6 | 29.6 | 36.4 | 25.2 | 33.3 |
83
+ | MathOctopus<sup>P</sup>-LoRA | 30.4 | 15.2 | 23.6 | 10.4 | 22.8 | 24.8 | 26.4 | 18.0 | 22.0 | 14.8 | 20.8 |
84
+ | MathOctopus<sup>P</sup> | 52.4 | 39.2 | 38.4 | 28.8 | 44.8 | 42.4 | 43.6 | 36.0 | 39.6 | 34.4 | 40.0 |
85
+ | **xRFT**-MathOctopus<sup>P</sup>| 54.8 | 38.4 | 45.2 | 33.2 | 43.6 | 45.2 | 38.0 | 35.6 | 48.4 | 36.4 | 41.9 |
86
+ <p></p >
87
+
88
+ | 13B Model | En | Sw | Zh | Bn | De | Es | Fr | Ja | Ru | Th | Overall |
89
+ |:--------------------------------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|
90
+ | MathOctopus<sup>C</sup> | 56.4 | 27.2 | 39.2 | 24.0 | 47.6 | 49.6 | 47.6 | 40.4 | 42.0 | 24.8 | 39.9 |
91
+ | **xRFT**-MathOctopus<sup>C</sup>| 53.6 | 28.0 | 45.2 | 21.2 | 48.0 | 46.4 | 46.0 | 35.2 | 45.6 | 28.8 | 39.8 |
92
+ | MathOctopus<sup>P</sup> | 53.2 | 42.8 | 48.8 | 35.2 | 44.4 | 48.0 | 48.4 | 43.2 | 47.6 | 46.8 | 45.8 |
93
+ | **xRFT**-MathOctopus<sup>P</sup>| 51.6 | 46.0 | 51.2 | 42.0 | 49.2 | 53.2 | 49.6 | 39.6 | 47.6 | 46.0 | 47.6 |
94
+ <p></p >
95
+
96
+ | 30-34B Model | En | Sw | Zh | Bn | De | Es | Fr | Ja | Ru | Th | Overall |
97
+ |:--------------------------------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|
98
+ | MathOctopus<sup>C</sup> | 55.6 | 24.4 | 36.0 | 19.2 | 40.4 | 51.2 | 44.4 | 27.2 | 37.2 | 21.6 | 35.7 |
99
+ | **xRFT**-MathOctopus<sup>C</sup>| 53.6 | 27.6 | 34.4 | 19.2 | 47.2 | 47.6 | 44.8 | 30.8 | 38.8 | 22.8 | 36.7 |
100
+ | MathOctopus<sup>P</sup> | 56.4 | 46.8 | 52.0 | 35.2 | 47.2 | 53.2 | 48.0 | 39.2 | 45.6 | 41.2 | 46.5 |
101
+ | **xRFT**-MathOctopus<sup>P</sup>| 51.6 | 47.2 | 52.4 | 37.6 | 51.2 | 52.8 | 44.4 | 41.6 | 50.0 | 47.6 | 47.6 |
102
+
103
+
104
+ ### **Overall Results on MSVAMP**
105
+
106
+ | 7B Model | En | Sw | Zh | Bn | De | Es | Fr | Ja | Ru | Th | Overall |
107
+ |:--------------------------------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|
108
+ | MathOctopus<sup>C</sup> | 49.2 | 36.6 | 43.6 | 30.2 | 48.6 | 46.8 | 46.4 | 42.5 | 46.7 | 34.0 | 42.5 |
109
+ | **xRFT**-MathOctopus<sup>C</sup>| 49.9 | 37.7 | 43.3 | 32.9 | 46.5 | 47.6 | 47.3 | 42.7 | 46.6 | 36.2 | 43.1 |
110
+ | MathOctopus<sup>P</sup>-LoRA | 30.4 | 15.2 | 23.6 | 10.4 | 22.8 | 24.8 | 26.4 | 18.0 | 22.0 | 14.8 | 20.8 |
111
+ | MathOctopus<sup>P</sup> | 46.5 | 40.1 | 42.5 | 29.1 | 43.5 | 45.4 | 46.0 | 42.5 | 45.4 | 35.7 | 41.7 |
112
+ | **xRFT**-MathOctopus<sup>P</sup>| 46.8 | 42.3 | 43.2 | 32.8 | 43.1 | 44.5 | 45.3 | 43.2 | 42.1 | 40.5 | 42.4 |
113
+ <p></p >
114
+
115
+ | 13B Model | En | Sw | Zh | Bn | De | Es | Fr | Ja | Ru | Th | Overall |
116
+ |:--------------------------------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|
117
+ | MathOctopus<sup>C</sup> | 56.6 | 40.4 | 49.0 | 30.3 | 50.9 | 54.2 | 54.7 | 46.3 | 52.4 | 35.7 | 47.1 |
118
+ | **xRFT**-MathOctopus<sup>C</sup>| 52.9 | 41.9 | 49.2 | 34.1 | 50.5 | 52.8 | 51.5 | 45.8 | 50.2 | 35.7 | 46.5 |
119
+ | MathOctopus<sup>P</sup> | 50.7 | 43.4 | 42.6 | 31.8 | 48.4 | 49.4 | 50.6 | 41.1 | 46.9 | 39.3 | 44.4 |
120
+ | **xRFT**-MathOctopus<sup>P</sup>| 44.6 | 43.4 | 46.4 | 34.2 | 47.7 | 48.2 | 49.9 | 43.1 | 48.2 | 39.5 | 44.5 |
121
+ <p></p >
122
+
123
+ | 30-34B Model | En | Sw | Zh | Bn | De | Es | Fr | Ja | Ru | Th | Overall |
124
+ |:--------------------------------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|:--------|
125
+ | MathOctopus<sup>C</sup> | 51.5 | 42.1 | 46.2 | 23.2 | 50.5 | 52.1 | 52.9 | 42.2 | 50.5 | 33.4 | 44.5 |
126
+ | **xRFT**-MathOctopus<sup>C</sup>| 48.1 | 42.8 | 43.6 | 23.3 | 48.7 | 50.0 | 48.9 | 43.4 | 44.6 | 35.5 | 42.9 |
127
+ | MathOctopus<sup>P</sup> | 56.4 | 46.8 | 52.0 | 35.2 | 47.2 | 53.2 | 48.0 | 39.2 | 45.6 | 41.2 | 46.5 |
128
+ | **xRFT**-MathOctopus<sup>P</sup>| 48.0 | 42.3 | 46.1 | 36.2 | 47.5 | 48.5 | 48.3 | 45.8 | 47.2 | 41.2 | 45.1 |
129
+
130
+
131
+ ### **MathOctopus in English**
132
+
133
+ | Models | GSM8K | SVAMP |
134
+ |:--------------------------------|:--------|:--------|
135
+ | LLaMA 2-7B | 42.4 | 38.3 |
136
+ | MathOctopus<sup>P</sup>-7B | 49.3 | 46.8 |
137
+ | MathOctopus<sup>C</sup>-7B | 50.8 | 49.3 |
138
+ | LLaMA 2-13B | 51.0 | 50.9 |
139
+ | MathOctopus<sup>P</sup>-13B | 55.5 | 52.1 |
140
+ | MathOctopus<sup>C</sup>-13B | 56.6 | 56.6 |
141
+ | LLaMA 1-33B | 50.0 | 49.0 |
142
+ | MathOctopus<sup>P</sup>-33B | 56.0 | 52.5 |
143
+ | MathOctopus<sup>C</sup>-33B | 53.7 | 51.5 |
144
+
145
+ ## Intended Uses
146
+ These models are trained for research purposes. They are designed to solve multilingual math problems. They can be used in educational software, tutoring systems, or any application where a solution to a math problem is needed.
147
+
148
+ ## Citation
149
+ Please cite our paper if you use our data, model or code. Please also kindly cite the original dataset papers.
150
+
151
+ ```
152
+ @misc{chen2023breaking,
153
+ title={Breaking Language Barriers in Multilingual Mathematical Reasoning: Insights and Observations},
154
+ author={Nuo Chen and Zinan Zheng and Ning Wu and Linjun Shou and Ming Gong and Yangqiu Song and Dongmei Zhang and Jia Li},
155
+ year={2023},
156
+ eprint={2310.20246},
157
+ archivePrefix={arXiv},
158
+ primaryClass={cs.CL}
159
+ }
160
+ ```