34 lines
1.5 KiB
Markdown
34 lines
1.5 KiB
Markdown
|
---
|
||
|
|
base_model:
|
||
|
|
- Qwen/Qwen3-1.7B
|
||
|
|
datasets:
|
||
|
|
- taki555/DeepScaleR-Easy
|
||
|
|
language:
|
||
|
|
- en
|
||
|
|
license: apache-2.0
|
||
|
|
pipeline_tag: text-generation
|
||
|
|
library_name: transformers
|
||
|
|
---
|
||
|
|
|
||
|
|
# Art-Qwen3-1.7B
|
||
|
|
|
||
|
|
This model is the Chain-of-Thought (CoT) efficient version of [Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B), developed as part of the research presented in the paper "[The Art of Efficient Reasoning: Data, Reward, and Optimization](https://huggingface.co/papers/2602.20945)".
|
||
|
|
|
||
|
|
## Model Description
|
||
|
|
|
||
|
|
Art-Qwen3-1.7B is optimized for efficient reasoning, aiming to produce short yet accurate thinking trajectories. It was trained using Reinforcement Learning (RL) with specialized reward shaping on the [DeepScaleR-Easy](https://huggingface.co/datasets/taki555/DeepScaleR-Easy) dataset. The training follows a two-stage paradigm involving length adaptation and reasoning refinement to maintain high accuracy while reducing computational overhead.
|
||
|
|
|
||
|
|
- **Paper:** [The Art of Efficient Reasoning: Data, Reward, and Optimization](https://huggingface.co/papers/2602.20945)
|
||
|
|
- **Project Page:** [https://wutaiqiang.github.io/project/Art](https://wutaiqiang.github.io/project/Art)
|
||
|
|
- **Base Model:** [Qwen/Qwen3-1.7B](https://huggingface.co/Qwen/Qwen3-1.7B)
|
||
|
|
|
||
|
|
## Citation
|
||
|
|
|
||
|
|
```bibtex
|
||
|
|
@inproceedings{wu2026art,
|
||
|
|
title={The Art of Efficient Reasoning: Data, Reward, and Optimization},
|
||
|
|
author={Taiqiang Wu and Zenan Xu and Bo Zhou and Ngai Wong},
|
||
|
|
year={2026},
|
||
|
|
url={https://arxiv.org/pdf/2602.20945}
|
||
|
|
}
|
||
|
|
```
|