Video Understanding on VideoMME (Subtitle Accuracy)
67.4Accuracy (No Subtitles)VisReflect
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| VisReflect2026.06 | 67.4 | 72.7 | — | — | — | — | |
| Qwen2.5-VL2026.06 | 65.2 | 70.7 | — | — | — | — | |
| VanillaBackbone=Qwen2.5-VL 7B, Token reduction ratio=0%2024.12 | 65.1 | 71.6 | 100 | 100 | — | — | |
| VideoChat-R12026.06 | 64.3 | 69.1 | — | — | — | — | |
| NVILA2026.06 | 64.2 | 70 | — | — | — | — | |
| LLaVA-VideoRelative token budget=100%, Prefilling FLOPs (T)=80.2, FLOPs Ratio=100%2026.04 | 64.1 | 71.4 | — | — | — | — | |
| iLLaVABackbone=Qwen2.5-VL 7B, Token reduction ratio=90%2024.12 | 63 | 69.9 | 97.7 | 96.7 | — | — | |
| iLLaVABackbone=Qwen2.5-VL 7B, Token reduction ratio=95%2024.12 | 62.3 | 68.6 | 96.6 | 95.7 | — | — | |
| HieraVidRelative token budget=30%, Prefilling FLOPs (T)=19.6, FLOPs Ratio=24.5%2026.04 | 62.3 | 70.8 | — | — | — | — | |
| VisionZipBackbone=Qwen2.5-VL 7B, Token reduction ratio=90%2024.12 | 62 | 69.2 | 96.6 | 95.3 | — | — | |
| Framefusion (ICCV2025)Relative token budget=30%, Prefilling FLOPs (T)=19.1, FLOPs Ratio=23.8%2026.04 | 61.9 | 70.1 | — | — | — | — | |
| HieraVidRelative token budget=20%, Prefilling FLOPs (T)=13.8, FLOPs Ratio=17.2%2026.04 | 61.8 | 66.3 | — | — | — | — | |
| FasterVLMBackbone=Qwen2.5-VL 7B, Token reduction ratio=90%2024.12 | 61.6 | 68.6 | 95.8 | 94.7 | — | — | |
| PruneVid (ACL2025)Relative token budget=30%, Prefilling FLOPs (T)=19.1, FLOPs Ratio=23.8%2026.04 | 61.5 | 69.6 | — | — | — | — | |
| PyramidDropBackbone=Qwen2.5-VL 7B, Token reduction ratio=90%2024.12 | 61.4 | 68.6 | 95.8 | 94.5 | — | — | |
| FasterVLMBackbone=Qwen2.5-VL 7B, Token reduction ratio=95%2024.12 | 60.9 | 67.8 | 94.8 | 93.8 | — | — | |
| VisionZipBackbone=Qwen2.5-VL 7B, Token reduction ratio=95%2024.12 | 60.9 | 67.9 | 94.9 | 93.6 | — | — | |
| PruMerge (ICCV2025)Relative token budget=30%, Prefilling FLOPs (T)=19.1, FLOPs Ratio=23.8%2026.04 | 60.9 | 67.7 | — | — | — | — | |
| LongVU2026.06 | 60.6 | 63.7 | — | — | — | — | |
| SparseVLMBackbone=Qwen2.5-VL 7B, Token reduction ratio=90%2024.12 | 60.3 | 67.1 | 93.7 | 92.9 | — | — | |
| PyramidDropBackbone=Qwen2.5-VL 7B, Token reduction ratio=95%2024.12 | 60.2 | 67.6 | 94.5 | 93.2 | — | — | |
| LongVILA2026.06 | 60.1 | 65.1 | — | — | — | — | |
| SparseVLMBackbone=Qwen2.5-VL 7B, Token reduction ratio=95%2024.12 | 59.6 | 66.4 | 92.8 | 91.6 | — | — | |
| Framefusion (ICCV2025)Relative token budget=20%, Prefilling FLOPs (T)=12.3, FLOPs Ratio=15.5%2026.04 | 59.5 | 66.2 | — | — | — | — | |
| FastV (ECCV2024)Relative token budget=30%, Prefilling FLOPs (T)=23.5, FLOPs Ratio=39.3%2026.04 | 59.3 | 66.7 | — | — | — | — | |
| PruneVid (ACL2025)Relative token budget=20%, Prefilling FLOPs (T)=12.3, FLOPs Ratio=15.5%2026.04 | 59.3 | 66.2 | — | — | — | — | |
| HieraVidRelative token budget=10%, Prefilling FLOPs (T)=6.85, FLOPs Ratio=8.5%2026.04 | 58.8 | 63.2 | — | — | — | — | |
| LLaVA-OneVisionRelative token budget=100%, Prefilling FLOPs (T)=40.8, FLOPs Ratio=100%2026.04 | 58.5 | 61.9 | — | — | — | — | |
| LLaVA-OneVision2026.06 | 58.2 | 61.5 | — | — | — | — | |
| HieraVidRelative token budget=30%, Prefilling FLOPs (T)=10.7, FLOPs Ratio=26.3%2026.04 | 58 | 61.6 | — | — | — | — | |
| PruneVid (ACL2025)Relative token budget=10%, Prefilling FLOPs (T)=6.19, FLOPs Ratio=7.7%2026.04 | 57.7 | 62.8 | — | — | — | — | |
| VanillaBase Model=Qwen2-VL, Relative Token Budget=100%, Prefilling FLOPs (T)=27.1, FLOPs Ratio=100.02026.04 | 57.6 | 61 | — | — | 59.3 | 100 | |
| FramefusionRelative token budget=30%, Prefilling FLOPs (T)=10.6, FLOPs Ratio=25.9%2026.04 | 57.5 | 61.1 | — | — | — | — | |
| PruneVidRelative token budget=30%, Prefilling FLOPs (T)=10.6, FLOPs Ratio=25.9%2026.04 | 57.4 | 59.9 | — | — | — | — | |
| Framefusion (ICCV2025)Relative token budget=10%, Prefilling FLOPs (T)=6.1, FLOPs Ratio=7.6%2026.04 | 57.3 | 61.7 | — | — | — | — | |
| HieraVidRelative token budget=20%, Prefilling FLOPs (T)=7.5, FLOPs Ratio=18.5%2026.04 | 57.1 | 60.7 | — | — | — | — | |
| PruMergeRelative token budget=30%, Prefilling FLOPs (T)=10.6, FLOPs Ratio=25.9%2026.04 | 56.6 | 58.8 | — | — | — | — | |
| PruMerge (ICCV2025)Relative token budget=20%, Prefilling FLOPs (T)=12.3, FLOPs Ratio=15.5%2026.04 | 56.5 | 62.6 | — | — | — | — | |
| FastVRelative token budget=30%, Prefilling FLOPs (T)=12.7, FLOPs Ratio=31.2%2026.04 | 56.3 | 58.1 | — | — | — | — | |
| HieraVidBase Model=Qwen2-VL, Relative Token Budget=30%, Prefilling FLOPs (T)=7.4, FLOPs Ratio=27.22026.04 | 56.1 | 60.1 | — | — | 58.1 | 98 | |
| PruneVidRelative token budget=20%, Prefilling FLOPs (T)=6.9, FLOPs Ratio=16.9%2026.04 | 55.8 | 58.4 | — | — | — | — | |
| FramefusionRelative token budget=20%, Prefilling FLOPs (T)=6.9, FLOPs Ratio=16.9%2026.04 | 55.7 | 60.1 | — | — | — | — | |
| StreamingLLM (ICLR2024)Relative token budget=30%, Prefilling FLOPs (T)=60.3, FLOPs Ratio=75.2%2026.04 | 55.5 | 65.7 | — | — | — | — | |
| FastV (ECCV2024)Relative token budget=20%, Prefilling FLOPs (T)=17.1, FLOPs Ratio=21.4%2026.04 | 55.4 | 61 | — | — | — | — | |
| FramefusionBase Model=Qwen2-VL, Relative Token Budget=30%, Prefilling FLOPs (T)=7.3, FLOPs Ratio=27.02026.04 | 55.4 | 59.7 | — | — | 57.6 | 97 | |
| HieraVidRelative token budget=10%, Prefilling FLOPs (T)=3.8, FLOPs Ratio=9.3%2026.04 | 55.4 | 59.7 | — | — | — | — | |
| PruMergeRelative token budget=20%, Prefilling FLOPs (T)=6.9, FLOPs Ratio=16.9%2026.04 | 54.2 | 56.5 | — | — | — | — | |
| PruneVidBase Model=Qwen2-VL, Relative Token Budget=30%, Prefilling FLOPs (T)=7.3, FLOPs Ratio=27.02026.04 | 53.8 | 57.3 | — | — | 55.6 | 93.7 | |
| FastVRelative token budget=20%, Prefilling FLOPs (T)=9.3, FLOPs Ratio=22.8%2026.04 | 53.7 | 57.1 | — | — | — | — | |
| PruneVidRelative token budget=10%, Prefilling FLOPs (T)=4.3, FLOPs Ratio=10.6%2026.04 | 53.7 | 57.5 | — | — | — | — | |
| FramefusionRelative token budget=10%, Prefilling FLOPs (T)=3.4, FLOPs Ratio=8.4%2026.04 | 53.5 | 56.5 | — | — | — | — | |
| StreamingLLMRelative token budget=30%, Prefilling FLOPs (T)=30.6, FLOPs Ratio=75%2026.04 | 52 | 54.9 | — | — | — | — | |
| PruMergeRelative token budget=10%, Prefilling FLOPs (T)=3.4, FLOPs Ratio=8.3%2026.04 | 51.9 | 54.7 | — | — | — | — | |
| FastVRelative token budget=10%, Prefilling FLOPs (T)=6.037, FLOPs Ratio=14.8%2026.04 | 51.7 | 55.1 | — | — | — | — | |
| StreamingLLM (ICLR2024)Relative token budget=20%, Prefilling FLOPs (T)=56.7, FLOPs Ratio=70.7%2026.04 | 51.6 | 61 | — | — | — | — | |
| PruMerge (ICCV2025)Relative token budget=10%, Prefilling FLOPs (T)=5.9, FLOPs Ratio=7.4%2026.04 | 50 | 55 | — | — | — | — | |
| FastV (ECCV2024)Relative token budget=10%, Prefilling FLOPs (T)=11.21, FLOPs Ratio=14.0%2026.04 | 49.6 | 52.5 | — | — | — | — | |
| StreamingLLMRelative token budget=20%, Prefilling FLOPs (T)=29.5, FLOPs Ratio=72.3%2026.04 | 49.5 | 52.7 | — | — | — | — | |
| StreamingLLMRelative token budget=10%, Prefilling FLOPs (T)=26.9, FLOPs Ratio=65.9%2026.04 | 48.8 | 51.2 | — | — | — | — | |
| StreamingLLM (ICLR2024)Relative token budget=10%, Prefilling FLOPs (T)=52.1, FLOPs Ratio=65.0%2026.04 | 45.8 | 53.9 | — | — | — | — | |
| DEViL2025.12 | 0.577 | 0.624 | — | — | — | — | |
| TimeMarker2025.12 | 0.573 | — | — | — | — | — | |
| VideoLLaMA2-7BParameters=7B2025.12 | 0.479 | — | — | — | — | — | |
| VideoChat2-7BParameters=7B2025.12 | 0.423 | 0.546 | — | — | — | — | |
| Video-LLaVA-7BParameters=7B2025.12 | 0.399 | 0.416 | — | — | — | — | |
| TimeChat-7BParameters=7B2025.12 | 0.347 | — | — | — | — | — | |
| LLaVA-NeXT-Video-7BParameters=7B2025.12 | 0.337 | — | — | — | — | — | |
| GPT-4oSubtitles=false2026.04 | — | — | — | — | 67.9 | — | |
| InternVL2.5-8BSubtitles=false2026.04 | — | — | — | — | 63.7 | — | |
| LLaVA-OneVision-7BSubtitles=false2026.04 | — | — | — | — | 58.2 | — | |
| LLaVA-Video-7BSubtitles=false2026.04 | — | — | — | — | 63.7 | — | |
| LongVILA-R1-7BFrames=162026.04 | — | — | — | — | 55.5 | — | |
| LongVILA-R1-7BFrames=322026.04 | — | — | — | — | 60.2 | — | |
| LongVILA-R1-7BFrames=642026.04 | — | — | — | — | 61.6 | — | |
| MiniCPM-V 2.6-8BSubtitles=false2026.04 | — | — | — | — | 59.7 | — | |
| MOSS-ChatVSubtitles=false2026.04 | — | — | — | — | 60 | — | |
| mPLUG-Owl3-8BSubtitles=false2026.04 | — | — | — | — | 53.5 | — | |
| Qwen2.5-VL-7BSubtitles=false2026.04 | — | — | — | — | 65.1 | — | |
| Qwen2.5-VL-7BFrames=162026.04 | — | — | — | — | 58.2 | — | |
| Qwen2.5-VL-7BFrames=322026.04 | — | — | — | — | 60.7 | — | |
| Qwen2.5-VL-7BFrames=642026.04 | — | — | — | — | 62.3 | — | |
| Qwen2.5-VL-7B-SFTFrames=162026.04 | — | — | — | — | 58.2 | — | |
| Qwen2.5-VL-7B-SFTFrames=322026.04 | — | — | — | — | 60.7 | — | |
| Qwen2.5-VL-7B-SFTFrames=642026.04 | — | — | — | — | 62.2 | — | |
| RLERSubtitles=false2026.04 | — | — | — | — | 68.5 | — | |
| STAR-R1Subtitles=false2026.04 | — | — | — | — | 56.6 | — | |
| TinyLLaVA-Video-R1Subtitles=false2026.04 | — | — | — | — | 46.6 | — | |
| TW-GRPOFrames=162026.04 | — | — | — | — | 58.2 | — | |
| TW-GRPOFrames=322026.04 | — | — | — | — | 61.2 | — | |
| TW-GRPOFrames=642026.04 | — | — | — | — | 62.7 | — | |
| Video-R1Subtitles=false2026.04 | — | — | — | — | 59.3 | — | |
| Video-R1Frames=162026.04 | — | — | — | — | 56.9 | — | |
| Video-R1Frames=322026.04 | — | — | — | — | 60.2 | — | |
| Video-R1Frames=642026.04 | — | — | — | — | 61.2 | — | |
| Video-RTSFrames=162026.04 | — | — | — | — | 58.7 | — | |
| Video-RTSFrames=322026.04 | — | — | — | — | 61.3 | — | |
| Video-RTSFrames=642026.04 | — | — | — | — | 62.9 | — | |
| VideoChat-R1Subtitles=false2026.04 | — | — | — | — | 58.8 | — | |
| VideoChat-R1.5Subtitles=false2026.04 | — | — | — | — | 67.1 | — | |
| VideoLLaMA 3-7BSubtitles=false2026.04 | — | — | — | — | 61 | — |