Publications at top-tier venues including ICLR, ICML, and NeurIPS
Two papers accepted to NeurIPS 2025
ICLR 2025: Proposed the first online reward-weighted fine-tuning framework for flow matching models with Wasserstein regularization, enabling self-evolution without human-collected data
ICLR 2023 (Oral Presentation): Introduced Learnable Behavior Control, breaking 24 Atari human world records with 100x less data and stable MoE self-evolution in RL
ICML 2022: Pioneered Generalized Data Distribution Iteration, providing the first theoretical justification for data optimization in RL
Selected as reviewer for ICML 2025, ICLR 2025, and NeurIPS 2024 (January 2025)