Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRs (2025 · 123 citations)
Why do some base language models gain a lot from reinforcement learning on verifiable tasks while others stall? The authors observe that Qwen-2.5-3B improves sharply on the Countdown arithmetic puzzle under a given RL setup, while Llama-3.2-3B, trained the same way, barely moves.
Generative Reward Models (2024 · 58 citations)
Reward models trained with the Bradley-Terry objective on human preference pairs score well on familiar data but generalize poorly to unseen tasks.
Towards System 2 Reasoning in LLMs: Learning How to Think With Meta Chain-of-Thought (2025 · 47 citations)
Large language models handle routine math well but still fail on harder problems. The authors argue that ordinary Chain-of-Thought (CoT) is part of the cause. Written solutions in training data show only the polished final derivation. They omit the hidden, non-linear work of trying options, checking them and…
Big-Math: A Large-Scale, High-Quality Math Dataset for Reinforcement Learning in Language Models (2025 · 43 citations)
Reinforcement learning for reasoning in LLMs needs math problems whose answers can be checked automatically.

Co-authors here

12 of 14 co-authors with a page, the most papers together first.

Cited by

Papers of our library cite their works 269 times (of the 149,439 papers whose bibliographies we read); 108 citing papers with a page here →.

Updated 3 November 2026