Update README

README.md

@@ -3,9 +3,6 @@ base_model: deepseek-ai/deepseek-math-7b-base
 tags:
 - alignment-handbook
 - generated_from_trainer
-datasets:
-- AI-MO/numina-problems-sft-v1.7-preproc
-- AI-MO/tora-chosen-v0.7
 widget:
   - example_title: Math problem
     messages:
@@ -25,13 +22,15 @@ model-index:
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
-<img src="https://cdn-uploads.huggingface.co/production/uploads/6200d0a443eb0913fa2df7cc/4xNbaeRi6HaAeo7UoRDZR.png" alt="Zephyr Logo" width="800" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
+<img src="https://huggingface.co/AI-MO/Numina-Math-7B/blob/main/thumbnail.png" alt="Numina Logo" width="800" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
 
-# Model Card for Numina-Math-7B
 
-Numina-Math is a series of language models that are trained to solve math problems using tool integrated reasoning.
-Numina-Math-7b won the first AI Math Olympiad, with a score of 29/50 on the public and private tests sets.
-This model is a fine-tuned version of [deepseek-ai/deepseek-math-7b-base](https://huggingface.co/deepseek-ai/deepseek-math-7b-base) with two stages of training, first on a dataset with 863k math question answer pairs and then on a dataset with 73k examples of multi-step synthetic generations using tool integrated reasoning.
+# Model Card for NuminaMath 7B
+
+NuminaMath is a series of language models that are trained to solve math problems using tool-integrated reasoning. NuminaMath 7B won the first progress prize of the [AI Math Olympiad (AIMO)](https://aimoprize.com), with a score of 29/50 on the public and private tests sets. This model is a fine-tuned version of [deepseek-ai/deepseek-math-7b-base](https://huggingface.co/deepseek-ai/deepseek-math-7b-base) with two stages of supervised fine-tuning:
+
+* **Stage 1:** fine-tune the base model on a large, diverse dataset of natural language math problems and solutions, where each solution is templated with Chain of Thought (CoT) to facilitate learning. 
+* **Stage 2:** fine-tune the model from Stage 1 on a synthetic dataset of tool-integrated reasoning, where each math problem is decomposed into a sequence of rationales, Python programs, and their outputs. Here we followed [Microsoft’s ToRA paper](https://arxiv.org/abs/2309.17452) and prompted GPT-4 to produce solutions in the ToRA format with code execution feedback. Fine-tuning on this data produces a reasoning agent that can solve mathematical problems via a mix of natural language reasoning and use of the Python REPL to compute intermediate results.
 
 
 ![image/png](https://cdn-uploads.huggingface.co/production/uploads/6200d0a443eb0913fa2df7cc/NyhBs_gzg40iwL995DO9L.png)
@@ -39,9 +38,9 @@ This model is a fine-tuned version of [deepseek-ai/deepseek-math-7b-base](https:
 
 ## Model description
 
-- **Model type:** A 7B parameter Math model fine-tuned in two stages of training, first on a dataset with 863k math question answer pairs and then on a dataset with 73k examples of multi-step synthetic generations using tool integrated reasoning.
+- **Model type:** A 7B parameter Math model fine-tuned in two stages of training, first on a dataset with math problem-solution pairs and then on a dataset with examples of multi-step synthetic generations using tool integrated reasoning.
 - **Language(s) (NLP):** Primarily English
-- **License:** MIT
+- **License:** Apacahe 2.0
 - **Finetuned from model:** [deepseek-ai/deepseek-math-7b-base](https://huggingface.co/deepseek-ai/deepseek-math-7b-base)
 
 ### Model Sources
@@ -49,7 +48,6 @@ This model is a fine-tuned version of [deepseek-ai/deepseek-math-7b-base](https:
 <!-- Provide the basic links for the model. -->
 
 - **Repository:** Coming soon to https://github.com/huggingface/alignment-handbook
-- **Demo:** https://huggingface.co/spaces/AI-MO/math-olympiad-solver
 
 ## Intended uses & limitations
 
@@ -76,18 +74,19 @@ gen_config = {
 
 outputs = pipe(prompt, **gen_config)
 text = outputs[0]["generated_text"]
-
 print(text)
-python_code = re.findall(r"```python(.*?)```", text, re.DOTALL)[0]
+
 # WARNING: This code will execute the python code in the string. We show this for eductional purposes only.
 # Please refer to our full pipeline for a safer way to execute code.
+python_code = re.findall(r"```python(.*?)```", text, re.DOTALL)[0]
 exec(python_code)
 
 ```
 ## Bias, Risks, and Limitations
 
 <!-- This section is meant to convey both technical and sociotechnical limitations. -->
-Numina-Math-7B was created to solve math problems, the model has not been aligned to preferences beyond the domain of solving math, and should not be used in a general chat setting.
+
+NuminaMath 7B was created to solve problems in the narrow domain of competition-level mathematics. As a result, the model should not be used for general chat applications. With greedy decoding, we find the model is capable of solving problems at the level of [AMC 12](https://artofproblemsolving.com/wiki/index.php/2023_AMC_12A_Problems), but often struggles generate a valid solution on harder problems at the AIME and Math Olympiad level.
 
 
 ## Training procedure
@@ -123,19 +122,4 @@ The following hyperparameters were used during training:
 - Transformers 4.40.1
 - Pytorch 2.3.1
 - Datasets 2.18.0
-- Tokenizers 0.19.1
-
-## Citation
-
-If you find Numina-Math useful in your work, please cite it with:
-
-```
-@misc{beeching2024numina-math,
-      title={Numina Math}, 
-      author={Edward Beeching and Lewis Tunstall and Roman Soletskyi and Kashif Rasul and Shengyi Huang and Jia Li},
-      year={2024},
-      publisher = {Hugging Face},
-      journal = {Hugging Face repository},
-      howpublished = {\url{https://huggingface.co/AI-MO/Numina-Math-7B}}
-}
-```
+- Tokenizers 0.19.1