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. available in [F16, q8_0, q6_K, q4_K_S]
71 Pulls Updated 7 weeks ago
Updated 7 weeks ago
7 weeks ago
2570713ff25f · 66GB
Readme
pipeline_tag: text-generation
inference: true
license: apache-2.0
datasets: - simplescaling/s1K-1.1
base_model: - Qwen/Qwen2.5-32B-Instruct
library_name: transformers
Model Summary
s1.1 is our sucessor of s1 with better reasoning performance by leveraging reasoning traces from r1 instead of Gemini.
- Logs: https://wandb.ai/hashimoto-group/o1/runs/m1ilia77/overview
- Repository: simplescaling/s1
- Paper: https://arxiv.org/abs/2501.19393
This model is a successor of s1-32B with slightly better performance. Thanks to Ryan Marten for helping generate r1 traces for s1K.
Use
The model usage is documented here.
Evaluation
| Metric | s1-32B | s1.1-32B | o1-preview | o1 | DeepSeek-R1 | DeepSeek-R1-Distill-Qwen-32B |
|---|---|---|---|---|---|---|
| # examples | 1K | 1K | ? | ? | >800K | 800K |
| AIME2024 | 56.7 | >56.7< | 40.0 | 74.4 | 79.8 | 72.6 |
| AIME2025 I | 26.7 | >60.0< | 37.5 | ? | 65.0 | 46.1 |
| MATH500 | 93.0 | >95.4< | 81.4 | 94.8 | 97.3 | 94.3 |
| GPQA-Diamond | 59.6 | >63.6< | 75.2 | 77.3 | 71.5 | 62.1 |
Note that s1-32B and s1.1-32B use budget forcing in this table; specifically ignoring end-of-thinking and appending “Wait” once or twice.