2026
UNIVERSAL AND EFFICIENT LOADING BALANCING FOR RL TRAINING OF LARGE MULTIMODAL MODELS
ICLR 2026poster
Reinforcement learning (RL) is crucial for aligning Vision-Language Models (VLMs), but its practical application is hampered by significant system-level bottlenecks. The typical RL pipeline, encompassing data loading, inference-based rollouts, and model updates, suffers from severe inefficiencies wh…