Text Generation
Safetensors
qwen2
conversational
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}, 
}
```