README.md

---
license: apache-2.0
datasets:
- dyyyyyyyy/ScaleQuest-Math
language:
- en
metrics:
- accuracy
library_name: transformers
pipeline_tag: text-generation
---
<p align="center"><h2 align="center">Unleashing Reasoning Capability of LLMs via Scalable Question Synthesis from Scratch</h2></p>

# Model Card for Qwen2-Math-7B-ScaleQuest 

<!-- Provide a quick summary of what the model is/does. -->

We introduce ScaleQuest, a scalable and novel data synthesis method that utilizes small-size open-source models to generate questions from scratch without the need for seed data with complex augmentation constraints.

* 📑 Project Page: [https://scalequest.github.io](https://scalequest.github.io/)
* 💻 Code: [https://github.com/yyDing1/ScaleQuest](https://github.com/yyDing1/ScaleQuest/)
* 📖 Paper: [Unleashing Reasoning Capability of LLMs via Scalable Question Synthesis from Scratch](https://arxiv.org/abs/2410.18693)
* 💾 Models in the 🤗 HuggingFace Hub: [ScaleQuest-Models](https://huggingface.co/collections/dyyyyyyyy/scalequest-670a7dc2623c91990f28913b)

<p align="center">
<img src="https://github.com/yyDing1/ScaleQuest/raw/main/img/results.png">
</p>

## Datasets & Models

Math Dataset: [link](https://huggingface.co/datasets/dyyyyyyyy/ScaleQuest-Math)

We release two question generator models and four problem-solving models.

| Model | Type | MATH | Olympiad Bench | 🤗 HuggingFace<br />Download Link |
| - | :-: | :-: | :-: | :-: |
| ScaleQuest-DeepSeekMath-7B-QGen | question generator | - | - | [link](https://huggingface.co/dyyyyyyyy/ScaleQuest-DeepSeekMath-7B-QGen)
| ScaleQuest-Qwen2-Math-7B-QGen | question generator | - | - | [link](https://huggingface.co/dyyyyyyyy/ScaleQuest-Qwen2-Math-7B-QGen)
| Mistral-7B-ScaleQuest | problem solver | 62.9 | 26.8 | [link](https://huggingface.co/dyyyyyyyy/Mistral-7B-ScaleQuest) |
| Llama3-8B-ScaleQuest | problem solver | 64.4 | 25.3 | [link](https://huggingface.co/dyyyyyyyy/Llama3-8B-ScaleQuest) |
| DeepSeekMath-7B-ScaleQuest | problem solver | 66.6 | 29.9 | [link](https://huggingface.co/dyyyyyyyy/DeepSeekMath-7B-ScaleQuest) |
| Qwen2-Math-7B-ScaleQuest | problem solver | 73.4 | 38.5 | [link](https://huggingface.co/dyyyyyyyy/Qwen2-Math-7B-ScaleQuest) |

## Demo usage

Below is an example using `Qwen2-Math-7B-ScaleQuest`
```python
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "dyyyyyyyy/Qwen2-Math-7B-ScaleQuest"

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

question = "Find the value of $x$ that satisfies the equation $4x+5 = 6x+7$."

sys_prompt = "<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n"
query_prompt = "<|im_start|>user" + "\n"
# {query}
prompt_after_query = "\n" + "Please reason step by step, and put your final answer within \\boxed{}.<|im_end|>" + "\n"
resp_prompt = "<|im_start|>assistant" + "\n"
prompt_before_resp = ""
# {resp}
delim = "<|im_end|>" + "\n"

prefix_prompt = f"{query_prompt}{question}{prompt_after_query}{resp_prompt}{prompt_before_resp}".rstrip(" ")
full_prompt = sys_prompt + delim.join([prefix_prompt])

# print(full_prompt)

inputs = tokenizer(full_prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=512, do_sample=False)
print(tokenizer.decode(outputs[0][len(inputs.input_ids[0]):], skip_special_tokens=True))

```

## Citation

```bibtex
@article{ding2024unleashing,
    title={Unleashing Reasoning Capability of LLMs via Scalable Question Synthesis from Scratch}, 
    author={Ding, Yuyang and Shi, Xinyu and Liang, Xiaobo and Li, Juntao and Zhu, Qiaoming and Zhang, Min},
    journal={https://arxiv.org/abs/2410.18693}, 
    year={2024}
}
```