update readme

README.md

@@ -8,7 +8,7 @@ tags:
 - math
 - reasoning
 ---
-<img src="llemma.jpg" width="400">
+<img src="llemma.png" width="400">
 
 [Zhangir Azerbayev](https://zhangir-azerbayev.github.io/), [Hailey Schoelkopf](https://github.com/haileyschoelkopf), [Keiran Paster](https://keirp.com), [Marco Dos Santos](https://github.com/dsantosmarco), [Stephen McAleer](https://www.andrew.cmu.edu/user/smcaleer/), [Albert Q. Jiang](https://albertqjiang.github.io/), [Jia Deng](https://www.cs.princeton.edu/~jiadeng/), [Stella Biderman](https://www.stellabiderman.com/), [Sean Welleck](https://wellecks.com/)
 
@@ -29,12 +29,12 @@ On chain-of-thought mathematics tasks, Llemma models outperform Llama-2, Code Ll
 | Model      | Size | GSM8k  | [OCW](https://openreview.net/forum?id=IFXTZERXdM7)   | MMLU-STEM | [SAT](https://huggingface.co/datasets/mcaleste/sat_multiple_choice_math_may_23)   | MATH  |
 |------------|------|--------|-------|-----------|-------|-------|
 | Llama 2    | 7B   | 11.8%  | 3.7%  | 29.9%     | 25%   | 3.2%  |
-| Code Llama | 7B   | 10.5%  | 4.4%  | 25.1%     | 9.4%  | 4.4%  |
-| LLEMMA     | 7B   | 36.4%  | 7.7%  | 37.7%     | 53.1% | 17.2% |
-| Minerva    | 8B   | 16.2%  | 7.7%  | 35.6%     | -     | 14.1% |
+| Code Llama | 7B   | 10.5%  | 4.4%  | 25.1%     | 9.4%  | 4.5%  |
+| LLEMMA     | 7B   | **36.4%**  | **7.7%**  | **37.7%**     | **53.1%** | **18.0%** |
+| Minerva    | 8B   | 16.2%  | **7.7%**  | 35.6%     | -     | 14.1% |
 |------------|------|--------|-------|-----------|-------|-------|
-| Code Llama | 34B  | 29.6%  | 7.0%  | 40.5%     | 40.6% | 11.9% |
-| LLEMMA     | 34B  | 51.5%  | 11.8% | 49.0%     | 71.9% | 24.1% |
+| Code Llama | 34B  | 29.6%  | 7.0%  | 40.5%     | 40.6% | 12.2% |
+| LLEMMA     | 34B  | **51.5%**  | **11.8%** | **49.0%**     | **71.9%** | **25.0%** |
 |------------|------|--------|-------|-----------|-------|-------|
 | Minerva    | 62B  | 52.4%  | 12.0% | 53.9%     | -     | 27.6% |
 | Minerva    | 540B | 58.8%  | 17.6% | 63.9%     | -     | 33.6% |
@@ -44,10 +44,10 @@ Further performance can be extracted by using majority voting:
 
 | Model   | Size | GSM8k maj@100 | OCW maj@100 | MMLU-STEM maj@16 | SAT maj@16 | MATH maj@256 |
 |---------|------|-------------|-----------|-----------------|-----------|------------|
-| LLEMMA  | 7B   | 54.0%       | 14.3%     | 49.9%           | 78.1%     | 32.0%      |
+| LLEMMA  | 7B   | 54.0%       | 14.3%     | 49.9%           | 78.1%     | **33.5**      |
 | Minerva | 8B   | 28.4%       | 12.5%     | 43.4%           | -         | 25.4%      |
 |---------|------|-------------|-----------|-----------------|-----------|------------|
-| LLEMMA  | 34B  | 69.3%       | 18.4%     | 59.7%           | 81.3%     | 41.0%      |
+| LLEMMA  | 34B  | 69.3%       | 18.4%     | 59.7%           | 81.3%     | **43.1%**      |
 |---------|------|-------------|-----------|-----------------|-----------|------------|
 | Minerva | 62B  | 68.5%       | 23.5%     | 63.5%           | -         | 43.4%      |
 | Minerva | 540B | 78.5%       | 30.8%     | 75.0%           | -         | 50.3%      |
@@ -55,5 +55,15 @@ Further performance can be extracted by using majority voting:
 ### Tool Use and Theorem Proving
 In addition to chain-of-thought reasoning, Llemma has strong capabilities in computational mathematics tasks. For tool use and formal theorem proving evaluations, see [our paper](#).
 
+### Citation
+```
+@article{azerbayev2023llemma,
+    title={Llemma: an open language model for mathematics},
+    author={Zhangir Azerbayev and Hailey Schoelkopf and Keiran Paster and Marco Dos Santos and Stephen McAleer and Albert Q. Jiang and Jia Deng and Stella Biderman and Sean Welleck},
+    eprint={xyz.xyz},
+    archivePrefix={arXiv}
+    year={2023}
+}
+```