Keeping the Evidence Chain: Semantic Evidence Allocation for Training-Free Token Pruning in Video Temporal Grounding
Open paper
Benchmark
Model
Rank
Results
moment-retrieval-on-charades-sta
SemVID (Qwen3-VL-4B, 25% visual-token retention)
#28
R@1 IoU=0.5: 61.08
R@1 IoU=0.7: 33.17
R@1 IoU=0.3: 76.91
mIoU: 52.31
Rank counts only results with a code link.