List your company
ResearchFormerly WizWand

Video Understanding on VideoMME v1.0 (test)

60.8ScoreFastV (ECCV2024) + DyToK (7B)
41.5646.55551.5556.545Dec 7, 2025
Updated 8mo ago

Evaluation results

MethodLinks
FastV (ECCV2024) + DyToK (7B)Token Retention Ratio=50%, Backbone=Qwen2.5-VL, Input Frames=32, Model Scale=7B2025.1260.8
Vanilla (TMLR)Token Retention Ratio=100%, Backbone=Qwen2.5-VL, Input Frames=322025.1260.7
FastV (ECCV2024) + DyToK (7B)Token Retention Ratio=75%, Backbone=Qwen2.5-VL, Input Frames=32, Model Scale=7B2025.1260.7
VisionZip† + DyToK (7B)Token Retention Ratio=50%, Backbone=Qwen2.5-VL, Input Frames=32, Model Scale=7B, Pooling Compatibility=true2025.1260.2
VisionZip† + DyToK (7B)Token Retention Ratio=25%, Backbone=Qwen2.5-VL, Input Frames=32, Model Scale=7B, Pooling Compatibility=true2025.1259.9
VisionZip†Token Retention Ratio=75%, Backbone=Qwen2.5-VL, Input Frames=32, Pooling Compatibility=true2025.1259.7
VisionZip†Token Retention Ratio=50%, Backbone=Qwen2.5-VL, Input Frames=32, Pooling Compatibility=true2025.1259.6
VisionZip† + DyToK (7B)Token Retention Ratio=75%, Backbone=Qwen2.5-VL, Input Frames=32, Model Scale=7B, Pooling Compatibility=true2025.1259.3
VisionZip†Token Retention Ratio=25%, Backbone=Qwen2.5-VL, Input Frames=32, Pooling Compatibility=true2025.1259.2
FastV (ECCV2024) + DyToK (7B)Token Retention Ratio=25%, Backbone=Qwen2.5-VL, Input Frames=32, Model Scale=7B2025.1258.8
FastV (ECCV2024)Token Retention Ratio=75%, Backbone=Qwen2.5-VL, Input Frames=322025.1258.7
VisionZip† + DyToK (7B)Token Retention Ratio=15%, Backbone=Qwen2.5-VL, Input Frames=32, Model Scale=7B, Pooling Compatibility=true2025.1258.4
VisionZip†Token Retention Ratio=15%, Backbone=Qwen2.5-VL, Input Frames=32, Pooling Compatibility=true2025.1258.3
VisionZip† + DyToK (7B)Token Retention Ratio=10%, Backbone=Qwen2.5-VL, Input Frames=32, Model Scale=7B, Pooling Compatibility=true2025.1257.1
FastV (ECCV2024) + DyToK (7B)Token Retention Ratio=15%, Backbone=Qwen2.5-VL, Input Frames=32, Model Scale=7B2025.1256.9
VisionZip†Token Retention Ratio=10%, Backbone=Qwen2.5-VL, Input Frames=32, Pooling Compatibility=true2025.1255.8
FastV (ECCV2024)Token Retention Ratio=50%, Backbone=Qwen2.5-VL, Input Frames=322025.1255.7
FastV (ECCV2024)Token Retention Ratio=25%, Backbone=Qwen2.5-VL, Input Frames=322025.1253.1
FastV (ECCV2024)Token Retention Ratio=15%, Backbone=Qwen2.5-VL, Input Frames=322025.1248.9
FastV (ECCV2024)Token Retention Ratio=10%, Backbone=Qwen2.5-VL, Input Frames=322025.1242.3
FastV (ECCV2024) + DyToK (7B)Token Retention Ratio=10%, Backbone=Qwen2.5-VL, Input Frames=32, Model Scale=7B2025.1242.3