Prior Knowledge Integration via LLM Encoding and Pseudo Event Regulation for Video Moment Retrieval

Benchmark Model Rank Results
highlight-detection-on-qvhighlightsLLMEPET#11mAP: 40.33Hit@1: 65.69
highlight-detection-on-youtube-highlightsLLMEPET#6mAP: 75.3
moment-retrieval-on-charades-staLLMEPET#16R@1 IoU=0.5: 58.31R@1 IoU=0.7: 36.49
moment-retrieval-on-qvhighlightsLLMEPET#15mAP: 44.05R@1 IoU=0.5: 66.73R@1 IoU=0.7: 49.94mAP@0.5: 65.76
natural-language-moment-retrieval-on-tacosLLMEPET#7R@1,IoU=0.3: 52.73R@1,IoU=0.5: 40.12R@1,IoU=0.7: 22.78mIoU: 36.55
video-grounding-on-qvhighlightsLLMEPET#3R@1,IoU=0.7: 49.94R@1,IoU=0.5: 66.73