Number it: Temporal Grounding Videos like Flipping Manga

Benchmark Model Rank Results
highlight-detection-on-qvhighlightsNumPro#10mAP: 40.54Hit@1: 70.71
moment-retrieval-on-charades-staLongVA-7B-DPO + NumPro-FT (extra data)#69R@1 IoU=0.5: 42.0R@1 IoU=0.7: 20.6R@1 IoU=0.3: 63.8mIoU: 41.4