File size: 1,026 Bytes
f0174f5 6876f06 20f81ad f0174f5 6876f06 f0174f5 6876f06 f0174f5 6876f06 cfafd1b f0174f5 6876f06 f0174f5 6876f06 f0174f5 6876f06 f0174f5 6876f06 6c7c60b 6876f06 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 |
---
pipeline_tag: text-generation
inference: true
license: apache-2.0
datasets:
- simplescaling/s1K
library_name: transformers
tags:
- vllm
---
# Model Summary
> s1 is a reasoning model finetuned from Qwen2.5-32B-Instruct on just 1,000 examples. It matches o1-preview & exhibits test-time scaling via budget forcing.
- **Repository:** [simplescaling/s1](https://github.com/simplescaling/s1)
- **Paper:** https://arxiv.org/abs/2501.19393
# Use
The model usage is documented [here](https://github.com/simplescaling/s1?tab=readme-ov-file#inference).
# Citation
```bibtex
@misc{muennighoff2025s1simpletesttimescaling,
title={s1: Simple test-time scaling},
author={Niklas Muennighoff and Zitong Yang and Weijia Shi and Xiang Lisa Li and Li Fei-Fei and Hannaneh Hajishirzi and Luke Zettlemoyer and Percy Liang and Emmanuel Candès and Tatsunori Hashimoto},
year={2025},
eprint={2501.19393},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2501.19393},
}
``` |