s1: Simple test-time scaling
Muennighoff et al., 2025, s1 (Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing)
Subscriber deck — free during early access
2 cards in “1 Introduction” · whole deck
1 Introduction
- What question does the s1 paper ask?What is the simplest recipe that gives both strong reasoning and test-time scaling (better accuracy with more thinking)? o1 had shown such scaling without revealing how, and replications such as DeepSeek-R1 used large-scale RL on many samples.
- Muennighoff et al., 2025, s1 (Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing), §1, p. 20275
- What is the s1 recipe, in two parts?(1) Supervised fine-tuning of Qwen2.5-32B-Instruct on just 1,000 carefully chosen questions with reasoning traces (s1K), which took 26 minutes on 16 H100s; (2) budget forcing at test time to control how long the model thinks.
- Muennighoff et al., 2025, s1 (Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing), §1, p. 20275
Source
- Niklas Muennighoff, Zitong Yang, Weijia Shi, Xiang Lisa Li, Li Fei-Fei, Hannaneh Hajishirzi, Luke Zettlemoyer, Percy Liang, Emmanuel Candès and Tatsunori Hashimoto. 2025. s1: Simple test-time scaling. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pp. 20275–20321. License: CC BY 4.0. Flashcards adapted from this work. doi:10.18653/v1/2025.emnlp-main.1025 https://aclanthology.org/2025.emnlp-main.1025/