Skip to content
NeurIPS 2026 planner

Poster · Sydney

Is the Importance Ratio Necessary for Stable Reinforcement Learning in LLMs?

When
Tue 8 Dec 2026, 10:00–13:00 AEDT
Where
Hall 1-4, ICC Sydney
Exhibition Centre, Level 1
Session
Sydney Poster Session 1
Authors
Shuibai Zhang, Junhyuck Kim, Gyeongman Kim, Jaewoong Cho
Organisations
University of Wisconsin-Madison, KRAFTON, KAIST
QRS classification
Reinforcement learning, Language models

Open in scheduleOfficial page