The Thinking Leakage in Reinforcement Learning for LLMs
Source
arxiv.org
Author
Wenxuan Zhu, Shuming Liu, Bang An, Aznaur Aliev, Bing Li, Yibo Yang, Bernard Ghanem
Date
Why it matters
If you RL-train a hybrid reasoning modelA model trained to think — generating extended internal reasoning before answering — trading time and tokens for accuracy on hard problems.Full definition → in no-think mode, reported gains may just be the base model's thinking mode resurfacing. Compare against THINK-mode accuracy before crediting the recipe.
Terms in this piece · Glossary
reasoning model — A model trained to think — generating extended internal reasoning before answering — trading time and tokens for accuracy on hard problems.
