Publications

Preprints

MLREF: Efficient Module Reuse for Reward Design in Reinforcement Learning via Large Language Models

Published in arXiv preprint, arXiv:2608.18827, 2026

MLREF evolves a persistent module pool for reward design in RL, reusing effective components across iterations and outperforming strong baselines by 25.2% in locomotion and 6.6% in manipulation.

Recommended citation: Chenglin Liu, Xun Wang, Ruishuo Chen, Zhuoran Li, and Longbo Huang. (2026). "MLREF: Efficient Module Reuse for Reward Design in Reinforcement Learning via Large Language Models." arXiv preprint arXiv:2608.18827.
Download Paper

Conference Papers

F2T: Fast Failover in LLM Training with Near-Zero Overhead State Management

Published in World Artificial Intelligence Conference Academic (WAICA), 2026

F2T leverages surplus network capacity to quickly save and load LLM training states, reducing recovery time by up to 98% and GPU utilization loss by up to 68%.

Recommended citation: Bohan Zhao, Yuanhong Wang, Chenglin Liu, Jiaqi Pan, Guang Yang, Ruitao Liu, Tingrui Zhang, Kai Luo, and Wei Xu. (2026). "F2T: Fast Failover in LLM Training with Near-Zero Overhead State Management." In Proceedings of the World Artificial Intelligence Conference Academic (WAICA 2026).