README.md

---
license: mit
configs:
- config_name: Difficulty Score
  data_files: Qwen2.5-Math-7B--deepscaler--difficulty.csv
- config_name: Response
  data_files: Qwen2.5-Math-7B--deepscaler.csv
task_categories:
- reinforcement-learning
---

# Difficulty Estimation on DeepScaleR

We annotate the entire [**DeepScaleR**](https://huggingface.co/agentica-org/DeepScaleR-1.5B-Preview) dataset with a **difficulty score** based on the performance of the [Qwen 2.5-MATH-7B](https://huggingface.co/Qwen/Qwen2.5-Math-7B) model. This provides an adaptive signal for curriculum construction and model evaluation.

**DeepScaleR** is a curated dataset of 40,000 reasoning-intensive problems used to train and evaluate reinforcement learning-based methods for large language models.

## Difficulty Scoring Method

Difficulty scores are estimated using the **Qwen 2.5-MATH-7B** model with the following generation settings:

- `temperature = 0.6`
- `top_p = 0.9`
- `max_tokens = 4096`
- Inference performed using [vLLM](https://github.com/vllm-project/vllm)
- Each problem is attempted **128 times**

The difficulty score `d_i` for each problem is computed as:

     d_i = 100 × (1 - (# successes / 128))

This approach balances the evaluation signal:
- A **strong model** would trivially solve easy problems, compressing the difficulty scale.
- A **weak model** would fail uniformly, providing poor resolution.
- Qwen 2.5-MATH-7B was selected for its **mid-range capabilities**, offering meaningful gradients across a wide spectrum of problems.

## Difficulty Estimation on Other Datasets

We also apply the same difficulty estimation procedure to the following datasets:

- [Open Reasoner Zero](https://huggingface.co/datasets/lime-nlp/orz_math_difficulty)
- [MATH](https://huggingface.co/datasets/lime-nlp/MATH_difficulty)
- [GSM8K](https://huggingface.co/datasets/lime-nlp/GSM8K_difficulty)

## 📬 Contact

For questions or feedback, feel free to reach out to [**Taiwei Shi**](https://maksimstw.github.io/) at [[email protected]](mailto:[email protected]).

## 📚 Citations

Github: https://github.com/uscnlp-lime/verl

If you find our dataset useful, please cite [Efficient Reinforcement Finetuning via Adaptive Curriculum Learning](https://huggingface.co/papers/2504.05520):

```bibtex
@misc{shi2025efficientreinforcementfinetuningadaptive,
  title={Efficient Reinforcement Finetuning via Adaptive Curriculum Learning}, 
  author={Taiwei Shi and Yiyang Wu and Linxin Song and Tianyi Zhou and Jieyu Zhao},
  year={2025},
  eprint={2504.05520},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  url={https://arxiv.org/abs/2504.05520}, 
}
```