Update pipeline tag, base model, and add description

This PR updates the model card with the correct `pipeline_tag` to reflect the model's question-answering capabilities. It also corrects the `base_model` to point to the correct repository, and adds a short description at the beginning for clarity. The `tags` field is also adjusted for better accuracy.

README.md

@@ -1,55 +1,56 @@
 ---
-license: apache-2.0
+base_model: YangLing0818/SuperCorrect-7B
+pipeline_tag: question-answering
 language:
 - en
+library_name: transformers
+license: apache-2.0
 metrics:
 - accuracy
-base_model: BitStarWalkin/SuperCorrect-7B
-library_name: transformers
 tags:
-- llama-cpp
-- gguf-my-repo
+- llama
+- qwen
+- mathematical-reasoning
+- gguf
 ---
 
-# Triangle104/SuperCorrect-7B-Q5_K_M-GGUF
-This model was converted to GGUF format from [`BitStarWalkin/SuperCorrect-7B`](https://huggingface.co/BitStarWalkin/SuperCorrect-7B) using llama.cpp via the ggml.ai's [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space.
-Refer to the [original model card](https://huggingface.co/BitStarWalkin/SuperCorrect-7B) for more details on the model.
+# SuperCorrect-7B: A Fine-Tuned LLM for Enhanced Mathematical Reasoning
+
+SuperCorrect-7B is a 7B parameter Large Language Model fine-tuned for improved mathematical reasoning and self-correction capabilities. It utilizes a two-stage framework incorporating hierarchical thought templates and cross-model collaborative direct preference optimization. This model significantly outperforms other 7B models on MATH and GSM8K benchmarks.
+
+This model was converted to GGUF format from [`YangLing0818/SuperCorrect-7B`](https://huggingface.co/YangLing0818/SuperCorrect-7B).
+Refer to the [original model card](https://huggingface.co/YangLing0818/SuperCorrect-7B) for more details on the model.
 
 ---
 Model details:
--
-SuperCorrect: Supervising and Correcting Language Models with Error-Driven Insights
+- SuperCorrect: Supervising and Correcting Language Models with Error-Driven Insights
 
     SuperCorrect: Supervising and Correcting Language Models with Error-Driven Insights Ling Yang*, Zhaochen Yu*, Tianjun Zhang, Minkai Xu, Joseph E. Gonzalez,Bin Cui, Shuicheng Yan
 
     Peking University, Skywork AI, UC Berkeley, Stanford University
 
 Introduction
--
-This repo provides the official implementation of SuperCorrect a novel two-stage fine-tuning method for improving both reasoning accuracy and self-correction ability for LLMs.
+- This repo provides the official implementation of SuperCorrect a novel two-stage fine-tuning method for improving both reasoning accuracy and self-correction ability for LLMs.
 
 Notably, our SupperCorrect-7B model significantly surpasses powerful DeepSeekMath-7B by 7.8%/5.3% and Qwen2.5-Math-7B by 15.1%/6.3% on MATH/GSM8K benchmarks, achieving new SOTA performance among all 7B models.
 🚨 Unlike other LLMs, we incorporate LLMs with our pre-defined hierarchical thought template ([Buffer of Thought (BoT)](https://github.com/YangLing0818/buffer-of-thought-llm)) to conduct more deliberate reasoning than conventional CoT. It should be noted that our evaluation methods relies on pure mathematical reasoning abilities of LLMs, instead of leverage other programming methods such as PoT and ToRA.
 
 Examples
--
-🚨 For more concise and clear presentation, we omit some XML tags.
+- 🚨 For more concise and clear presentation, we omit some XML tags.
 Model details
 
 You can check our Github repo for more details.
 
 Quick Start
--
-Requirements
--
+- Requirements
     Since our current model is based on Qwen2.5-Math series, transformers>=4.37.0 is needed for Qwen2.5-Math models. The latest version is recommended.
 
     🚨 This is a must because `transformers` integrated Qwen2 codes since `4.37.0`.
 
 Inference
--
-🤗 Hugging Face Transformers
+- 🤗 Hugging Face Transformers
 
+```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
 model_name = "BitStarWalkin/SuperCorrect-7B"
@@ -62,7 +63,7 @@ model = AutoModelForCausalLM.from_pretrained(
 )
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 
-prompt = "Find the distance between the foci of the ellipse \[9x^2 + \frac{y^2}{9} = 99.\]"
+prompt = "Find the distance between the foci of the ellipse \\[9x^2 + \\frac{y^2}{9} = 99.\\]"
 hierarchical_prompt = "Solve the following math problem in a step-by-step XML format, each step should be enclosed within tags like <Step1></Step1>. For each step enclosed within the tags, determine if this step is challenging and tricky, if so, add detailed explanation and analysis enclosed within <Key> </Key> in this step, as helpful annotations to help you thinking and remind yourself how to conduct reasoning correctly. After all the reasoning steps, summarize the common solution and reasoning steps to help you and your classmates who are not good at math generalize to similar problems within <Generalized></Generalized>. Finally present the final answer within <Answer> </Answer>."
 # HT
 messages = [
@@ -87,67 +88,128 @@ generated_ids = [
 
 response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
 print(response)
+```
 
-Performance
--
-We evaluate our SupperCorrect-7B on two widely used English math benchmarks GSM8K and MATH. All evaluations are tested with our evaluation method which is zero-shot hierarchical thought based prompting.
+#### 🔥 vLLM
 
-Citation
--
-@article{yang2024supercorrect,
-title={SuperCorrect: Supervising and Correcting Language Models with Error-Driven Insights}
-  author={Yang, Ling and Yu, Zhaochen and Zhang, Tianjun and Xu, Minkai and Gonzalez, Joseph E and Cui, Bin and Yan, Shuicheng},
-  journal={arXiv preprint arXiv:2410.09008},
-  year={2024}
-}
-@article{yang2024buffer,
-  title={Buffer of Thoughts: Thought-Augmented Reasoning with Large Language Models},
-  author={Yang, Ling and Yu, Zhaochen and Zhang, Tianjun and Cao, Shiyi and Xu, Minkai and Zhang, Wentao and Gonzalez, Joseph E and Cui, Bin},
-  journal={arXiv preprint arXiv:2406.04271},
-  year={2024}
-}
+```python
+import os
+from vllm import LLM, SamplingParams
+model_name = 'BitStarWalkin/SuperCorrect-7B'
+hierarchical_prompt = "Solve the following math problem in a step-by-step XML format, each step should be enclosed within tags like <Step1></Step1>. For each step enclosed within the tags, determine if this step is challenging and tricky, if so, add detailed explanation and analysis enclosed within <Key> </Key> in this step, as helpful annotations to help you thinking and remind yourself how to conduct reasoning correctly. After all the reasoning steps, summarize the common solution and reasoning steps to help you and your classmates who are not good at math generalize to similar problems within <Generalized></Generalized>. Finally present the final answer within <Answer> </Answer>."
+prompts = [
+    "For what positive value of $t$ is $|{-4+ti}| = 6$?",
+    "Find the distance between the foci of the ellipse \\[9x^2 + \\frac{y^2}{9} = 99.\\]",
+    "The fourth term of a geometric series is $24$ and the eleventh term is $3072$. What is the common ratio?"
+]
+combined_prompts = [hierarchial_prompt + '\n' + prompt for prompt in prompts]
+sampling_params = SamplingParams(temperature=0, top_p=1,max_tokens=1024)
+llm = LLM(model=model_name, trust_remote_code=True)
+outputs = llm.generate(combined_prompts, sampling_params)
+
+#Print the outputs.
+for output in outputs:
+    prompt = output.prompt
+    generated_text = output.outputs[0].text
+    print(f"Prompt: {prompt}")
+    print(f"Generated text: {generated_text}")
+```
 
-Acknowledgements
--
-Our SuperCorrect is a two-stage fine-tuning model which based on several extraordinary open-source models like Qwen2.5-Math, DeepSeek-Math, Llama3-Series. Our evaluation method is based on the code base of outstanding works like Qwen2.5-Math and lm-evaluation-harness. We also want to express our gratitude for amazing works such as BoT which provides the idea of thought template.
+Here we also provide inference code with [vLLM](https://github.com/vllm-project/vllm) . vLLM is a fast and easy-to-use library for LLM inference and serving.
 
----
-## Use with llama.cpp
-Install llama.cpp through brew (works on Mac and Linux)
 
-```bash
-brew install llama.cpp
+### 1. Our evaluation  
 
-```
-Invoke the llama.cpp server or the CLI.
+Here we provide two different evaluation methods: **online version**  which utilizes GPT-4o to conduct a more fair and robust judgement and **offline version**  which utilizes programming method to verify the final results. Both methods aim to provide a more accurate and strict evaluation results, as the final results in MATH dataset are not always numeric or pure expression. We now provide online version for evaluation, we will update soon for offline version.
 
-### CLI:
 ```bash
-llama-cli --hf-repo Triangle104/SuperCorrect-7B-Q5_K_M-GGUF --hf-file supercorrect-7b-q5_k_m.gguf -p "The meaning to life and the universe is"
+API_KEY= "Input your key here"
+MODEL_NAME_OR_PATH="BitStarWalkin/SuperCorrect-7B"
+export CUDA_VISIBLE_DEVICES="0"
+bash evaluation.sh $API_KEY $MODEL_NAME_OR_PATH
 ```
 
-### Server:
+### 2. Evaluation with [lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness)
+
 ```bash
-llama-server --hf-repo Triangle104/SuperCorrect-7B-Q5_K_M-GGUF --hf-file supercorrect-7b-q5_k_m.gguf -c 2048
+lm_eval --model hf \
+    --model_args pretrained="Qwen2.5-Math-7B-Instruct" \
+    --tasks minerva_math \
+    --log_samples \
+    --output_path Qwen2.5-Math-7B-Instruct-lm-evaluation \
+    --batch_size 12
+
+lm_eval --model hf \
+    --model_args pretrained="SuperCorrect-7B" \
+    --tasks minerva_math \
+    --log_samples \
+    --output_path SuperCorrect-7B-lm-evaluation \
+    --batch_size 12
 ```
+Evaluation results produced by lm-evaluation:
+
+| Qwen2.5-Math-7B-Instruct    | Version | Filter | n-shot | Metric      |      |  Value |      | Stderr |
+| ----------------------------------- | ------: | ------ | -----: | ----------- | ---- | -----: | ---- | -----: |
+| minerva_math                        |       1 | none   |      4 | exact_match | ↑    | 0.5034 | ±    | 0.0064 |
+| - minerva_math_algebra              |       1 | none   |      4 | exact_match | ↑    | 0.7009 | ±    | 0.0133 |
+| - minerva_math_counting_and_prob    |       1 | none   |      4 | exact_match | ↑    | 0.5232 | ±    | 0.0230 |
+| - minerva_math_geometry             |       1 | none   |      4 | exact_match | ↑    | 0.4635 | ±    | 0.0228 |
+| - minerva_math_intermediate_algebra |       1 | none   |      4 | exact_match | ↑    | 0.2237 | ±    | 0.0139 |
+| - minerva_math_num_theory           |       1 | none   |      4 | exact_match | ↑    | 0.4667 | ±    | 0.0215 |
+| - minerva_math_prealgebra           |       1 | none   |      4 | exact_match | ↑    | 0.7394 | ±    | 0.0149 |
+| - minerva_math_precalc              |       1 | none   |      4 | exact_match | ↑    | 0.2143 | ±    | 0.0176 |
+
+| SuperCorrect-7B              | Version | Filter | n-shot | Metric      |      |  Value |      | Stderr |
+| ------------------------------------ | ------: | ------ | -----: | ----------- | ---- | -----: | ---- | -----: |
+| minerva_math                         |       1 | none   |      4 | exact_match | ↑    | 0.6188 (**+0.1154**) | ±    | 0.0065 |
+| - minerva_math_algebra               |       1 | none   |      4 | exact_match | ↑    | 0.7936 (**+0.0927**) | ±    | 0.0118 |
+| - minerva_math_counting_and_prob     |       1 | none   |      4 | exact_match | ↑    | 0.5802 (**+0.0570**) | ±    | 0.0227 |
+| - minerva_math_geometry              |       1 | none   |      4 | exact_match | ↑    | 0.5261 (**+0.0626**) | ±    | 0.0228 |
+| - minerva_math_intermediate_algebra  |       1 | none   |      4 | exact_match | ↑    | 0.4385 (**+0.2148**) | ±    | 0.0165 |
+| - minerva_math_num_theory            |       1 | none   |      4 | exact_match | ↑    | 0.6167 (**+0.1500**) | ±    | 0.0209 |
+| - minerva_math_prealgebra            |       1 | none   |      4 | exact_match | ↑    | 0.7715 (**+0.0321**) | ±    | 0.0142 |
+| - minerva_math_precalc               |       1 | none   |      4 | exact_match | ↑    | 0.4103 (**+0.1960**) | ±    | 0.0211 |
+
+| Summary       | Version | Filter | n-shot | Metric      |      |  Value |      | Stderr |
+| ------------ | ------: | ------ | ------ | ----------- | ---- | -----: | ---- | -----: |
+| Qwen2.5-Math-7B-Instruct |       1 | none   |       4| exact_match | ↑    | 0.5034 | ±    | 0.0064 |
+| SuperCorrect-7B  |       1 | none   |       4| exact_match | ↑    | 0.6188 (**+0.1154**) | ±    | 0.0065 |
+
+### 3. Evaluation with [Qwen2.5-Math-Evaluation](https://github.com/QwenLM/Qwen2.5-Math)
+```bash
+export CUDA_VISIBLE_DEVICES="0"
+MODEL_NAME_OR_PATH="Qwen/Qwen2.5-Math-7B-Instruct"
+bash sh/eval.sh $PROMPT_TYPE $MODEL_NAME_OR_PATH
 
-Note: You can also use this checkpoint directly through the [usage steps](https://github.com/ggerganov/llama.cpp?tab=readme-ov-file#usage) listed in the Llama.cpp repo as well.
-
-Step 1: Clone llama.cpp from GitHub.
-```
-git clone https://github.com/ggerganov/llama.cpp
+export CUDA_VISIBLE_DEVICES="0"
+MODEL_NAME_OR_PATH="BitStarWalkin/SuperCorrect-7B"
+bash sh/eval.sh $PROMPT_TYPE $MODEL_NAME_OR_PATH
 ```
+Evaluation results produced by Qwen2.5-Math-Eval:
+| Model            | MATH Accuracy (%) |
+| ---------------- | ----------------- |
+| Qwen2.5-Math     | 80.6              |
+| **SuperCorrect**     | **82.1**              |
+| **Our Improvement**   | **+1.5**          |
 
-Step 2: Move into the llama.cpp folder and build it with `LLAMA_CURL=1` flag along with other hardware-specific flags (for ex: LLAMA_CUDA=1 for Nvidia GPUs on Linux).
-```
-cd llama.cpp && LLAMA_CURL=1 make
-```
+## Citation
 
-Step 3: Run inference through the main binary.
-```
-./llama-cli --hf-repo Triangle104/SuperCorrect-7B-Q5_K_M-GGUF --hf-file supercorrect-7b-q5_k_m.gguf -p "The meaning to life and the universe is"
-```
-or 
-```
-./llama-server --hf-repo Triangle104/SuperCorrect-7B-Q5_K_M-GGUF --hf-file supercorrect-7b-q5_k_m.gguf -c 2048
+```bash
+@inproceedings{yang2025supercorrect,
+  title={SuperCorrect: Supervising and Correcting Language Models with Error-Driven Insights},
+  author={Yang, Ling and Yu, Zhaochen and Zhang, Tianjun and Xu, Minkai and Gonzalez, Joseph E and Cui, Bin and Yan, Shuicheng},
+  booktitle={International Conference on Learning Representations},
+  year={2025}
+}
+
+@article{yang2024buffer,
+  title={Buffer of Thoughts: Thought-Augmented Reasoning with Large Language Models},
+  author={Yang, Ling and Yu, Zhaochen and Zhang, Tianjun and Cao, Shiyi and Xu, Minkai and Zhang, Wentao and Gonzalez, Joseph E and Cui, Bin},
+  journal={Advances in Neural Information Processing Systems},
+  year={2024}
+}
 ```
+
+## Acknowledgements
+
+Our SuperCorrect is a two-stage fine-tuning model which based on several extraordinary open-source models like [Qwen2.5-Math](https://github.com/QwenLM/Qwen2.5-Math), [DeepSeek-Math](https://github.com/deepseek-ai/DeepSeek-Math), [Llama3-Series](https://github.com/meta-llama/llama3). Our evaluation method is based on the code base of outstanding works like [Qwen2.5-Math](https://github.com/QwenLM/Qwen2.5-Math) and  [lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness). We also want to express our gratitude for amazing works such as [BoT](https://github.com/YangLing0818/buffer-of-thought-llm) which provides the idea of thought template.