Leo Dianbo Liu (刘钿渤)

Leo Dianbo Liu (刘钿渤)

Developing fundamental machine learning to enable AI to conduct autonomous research

Expected Return Causes Outcome-Level Mode Collapse in Reinforcement Learning and How to Fix It with Inverse Probability Scaling

Published in ICML 2026, 2026

Expected Return Causes Outcome-Level Mode Collapse in Reinforcement Learning and How to Fix It with Inverse Probability Scaling.