Reinforcement Learning Tuning for VideoLLMs: Reward Design and Data Efficiency

Benchmark Model Rank Results
temporal-sentence-grounding-on-activitynet-captions-temporal-localization-val-2Temporal-RLT (4096 video tokens) (extra data)#5mIoU: 41.3
moment-retrieval-on-charades-staTemporal-RLT (4096 video tokens) (extra data)#72mIoU: 58.0