Multimodal Benchmarking on MMBench English
70.4AccuracyFSR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FSRBackbone=LLaVA-NeXT-13B, Tokens Retained=960, Pruning Ratio=66.7%2026.02 | 70.4 | — | |
| FMVR-LLaVA#Vision Tokens=576, Backbone=LLaVA-1.5-13B2026.03 | 69.5 | — | |
| CDPrunerBackbone=LLaVA-NeXT-13B, Tokens Retained=960, Pruning Ratio=66.7%2026.02 | 69.2 | — | |
| FSRBackbone=LLaVA-NeXT-13B, Tokens Retained=640, Pruning Ratio=77.8%2026.02 | 69.2 | — | |
| VisPrunerBackbone=LLaVA-NeXT-13B, Tokens Retained=960, Pruning Ratio=66.7%2026.02 | 69 | — | |
| PPLLAVAResolution=336*3362024.11 | 68.9 | — | |
| CDPrunerBackbone=LLaVA-NeXT-13B, Tokens Retained=640, Pruning Ratio=77.8%2026.02 | 68.9 | — | |
| LLaVA-NeXT-13BBackbone=LLaVA-NeXT-13B, Tokens Retained=2880, Pruning Ratio=0%2026.02 | 68.6 | — | |
| M3#Vision Tokens=1802026.03 | 68.6 | — | |
| LLaVA-1.5-13B#Vision Tokens=576, Backbone=LLaVA-1.5-13B2026.03 | 68.5 | — | |
| SparseVLM#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 68.4 | — | |
| FMVR-LLaVA#Vision Tokens=144, Backbone=LLaVA-1.5-13B2026.03 | 68.3 | — | |
| VisionZipBackbone=LLaVA-NeXT-13B, Tokens Retained=640, Pruning Ratio=77.8%2026.02 | 68.1 | — | |
| HoloVBackbone=LLaVA-NeXT-13B, Tokens Retained=320, Pruning Ratio=88.9%2026.02 | 67.9 | — | |
| FSRBackbone=LLaVA-NeXT-13B, Tokens Retained=320, Pruning Ratio=88.9%2026.02 | 67.9 | — | |
| LLaVA-1.5-13BResolution=336*3362024.11 | 67.8 | — | |
| VisPrunerBackbone=LLaVA-NeXT-13B, Tokens Retained=640, Pruning Ratio=77.8%2026.02 | 67.7 | — | |
| RubiCap-7B_PixMoCapAnnotator=RubiCap, Model Scale=7B, Caption Generation Strategy=PixMoCap2026.03 | 67.61 | — | |
| DivPruneBackbone=LLaVA-NeXT-13B, Tokens Retained=640, Pruning Ratio=77.8%2026.02 | 67.5 | — | |
| PyramidDrop#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 67.5 | — | |
| CDPruner#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 67.5 | — | |
| LLaVA-Next-7BResolution=672*6722024.11 | 67.4 | — | |
| VisionZip#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 67.4 | — | |
| M3#Vision Tokens=144, Backbone=LLaVA-1.5-13B2026.03 | 67.2 | — | |
| FMVR-LLaVA#Vision Tokens=28802026.03 | 67.1 | — | |
| TRIM#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 67.1 | — | |
| HoloVBackbone=LLaVA-NeXT-13B, Tokens Retained=640, Pruning Ratio=77.8%2026.02 | 67 | — | |
| CDPrunerRetained Tokens=960, Reduction Ratio=↓ 66.7%2026.02 | 66.9 | — | |
| FSRRetained Tokens=960, Reduction Ratio=↓ 66.7%2026.02 | 66.9 | — | |
| SPHINX#Vision Tokens=2892026.03 | 66.9 | — | |
| CDPrunerBackbone=LLaVA-NeXT-13B, Tokens Retained=320, Pruning Ratio=88.9%2026.02 | 66.8 | — | |
| VisionZipBackbone=LLaVA-NeXT-13B, Tokens Retained=320, Pruning Ratio=88.9%2026.02 | 66.5 | — | |
| VisPrunerBackbone=LLaVA-NeXT-13B, Tokens Retained=320, Pruning Ratio=88.9%2026.02 | 66.5 | — | |
| FMVR-LLaVA#Vision Tokens=7202026.03 | 66.5 | — | |
| MQT-LLaVA#Vision Tokens=144, Backbone=LLaVA-1.5-13B2026.03 | 66.5 | — | |
| M3#Vision Tokens=1442026.03 | 66.4 | — | |
| VisPruner#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 66.4 | — | |
| FSRRetained Tokens=640, Reduction Ratio=↓ 77.8%2026.02 | 66.3 | — | |
| Prumerge+#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 66.3 | — | |
| DivPrune#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 66.3 | — | |
| FMVR-LLaVA#Vision Tokens=36, Backbone=LLaVA-1.5-13B2026.03 | 66.3 | — | |
| CDPrunerRetained Tokens=640, Reduction Ratio=↓ 77.8%2026.02 | 66.2 | — | |
| FastV#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 66.1 | — | |
| DivPruneBackbone=LLaVA-NeXT-13B, Tokens Retained=320, Pruning Ratio=88.9%2026.02 | 65.9 | — | |
| LLaVA-NeXT-7BRetained Tokens=2880, Reduction Ratio=100%2026.02 | 65.9 | — | |
| FMVR-LLaVA#Vision Tokens=1802026.03 | 65.9 | — | |
| FMVR-LLaVA#Vision Tokens=5762026.03 | 65.9 | — | |
| VisPrunerRetained Tokens=960, Reduction Ratio=↓ 66.7%2026.02 | 65.8 | — | |
| DivPruneRetained Tokens=640, Reduction Ratio=↓ 77.8%2026.02 | 65.8 | — | |
| LLaVA-NeXT-7B#Vision Tokens=28802026.03 | 65.8 | — | |
| FMVR-LLaVA#Vision Tokens=1442026.03 | 65.8 | — | |
| RubiCap-3B_PixMoCapAnnotator=RubiCap, Model Scale=3B, Caption Generation Strategy=PixMoCap2026.03 | 65.72 | — | |
| FastVBackbone=LLaVA-NeXT-13B, Tokens Retained=640, Pruning Ratio=77.8%2026.02 | 65.5 | — | |
| HoloVRetained Tokens=320, Reduction Ratio=↓ 88.9%2026.02 | 65.5 | — | |
| CDPruner#Vision Tokens=3202026.03 | 65.5 | — | |
| CDPrunerRetained Tokens=320, Reduction Ratio=↓ 88.9%2026.02 | 65.4 | — | |
| DART#Vision Tokens=128, Backbone=LLaVA-1.5-13B2026.03 | 65.4 | — | |
| FMVR-LLaVA#Vision Tokens=362026.03 | 65.2 | — | |
| FMVR-LLaVA#Vision Tokens=9, Backbone=LLaVA-1.5-13B2026.03 | 65.1 | — | |
| VisPrunerRetained Tokens=640, Reduction Ratio=↓ 77.8%2026.02 | 64.9 | — | |
| FSRRetained Tokens=320, Reduction Ratio=↓ 88.9%2026.02 | 64.9 | — | |
| Prumerge+#Vision Tokens=1442026.03 | 64.9 | — | |
| M3Reduction Ratio=75%, Vision Tokens=362026.02 | 64.8 | — | |
| LLaVA-Next-VideoResolution=336*3362024.11 | 64.7 | — | |
| mPLUG-OwL2#Vision Tokens=10242026.03 | 64.5 | — | |
| RubiCap-3B_DenseFusionAnnotator=RubiCap, Model Scale=3B, Caption Generation Strategy=DenseFusion2026.03 | 64.43 | — | |
| HoloVBackbone=LLaVA-NeXT-13B, Tokens Retained=960, Pruning Ratio=66.7%2026.02 | 64.4 | — | |
| MQT-LLaVA#Vision Tokens=1442026.03 | 64.4 | — | |
| LLaVA-1.5-7BReduction Ratio=0%, Vision Tokens=5762026.02 | 64.3 | — | |
| HoloVRetained Tokens=640, Reduction Ratio=↓ 77.8%2026.02 | 64.3 | — | |
| LLaVA-v1.5#Vision Tokens=5762026.03 | 64.3 | — | |
| DART#Vision Tokens=3202026.03 | 64.2 | — | |
| FMVR-LLaVA#Vision Tokens=92026.03 | 64.2 | — | |
| DivPruneRetained Tokens=320, Reduction Ratio=↓ 88.9%2026.02 | 63.9 | — | |
| DivPrune#Vision Tokens=3202026.03 | 63.9 | — | |
| VisPrunerRetained Tokens=320, Reduction Ratio=↓ 88.9%2026.02 | 63.8 | — | |
| GPT-4VAnnotator=GPT-4V2026.03 | 63.57 | — | |
| MQTReduction Ratio=88%, Vision Tokens=642026.02 | 63.5 | — | |
| TRIM#Vision Tokens=3202026.03 | 63.5 | — | |
| MQTReduction Ratio=88%, Vision Tokens=362026.02 | 63.4 | — | |
| VisionZip#Vision Tokens=3202026.03 | 63.3 | — | |
| Mask-LLaVAReduction Ratio=97%, Vision Tokens=422026.02 | 63.2 | — | |
| PyramidDrop#Vision Tokens=1922026.03 | 63.2 | — | |
| Mask-LLaVAReduction Ratio=90%, Vision Tokens=572026.02 | 63.1 | — | |
| M3Reduction Ratio=75%, Vision Tokens=92026.02 | 63.1 | — | |
| FastVRetained Tokens=640, Reduction Ratio=↓ 77.8%2026.02 | 63.1 | — | |
| SparseVLM#Vision Tokens=3202026.03 | 63.1 | — | |
| CDPruner#Vision Tokens=1282026.03 | 63.1 | — | |
| Prumerge+#Vision Tokens=3202026.03 | 63 | — | |
| TRIM#Vision Tokens=1282026.03 | 63 | — | |
| VisionZip#Vision Tokens=1922026.03 | 63 | — | |
| Chat-Univ-1.5Resolution=336*3362024.11 | 62.7 | — | |
| VisPruner#Vision Tokens=1282026.03 | 62.7 | — | |
| SparseVLM#Vision Tokens=1922026.03 | 62.5 | — | |
| FMVR-LLaVA#Vision Tokens=52026.03 | 62.3 | — | |
| FMVR-LLaVA#Vision Tokens=1, Backbone=LLaVA-1.5-13B2026.03 | 62.2 | — | |
| Mask-LLaVAReduction Ratio=97%, Vision Tokens=152026.02 | 62.1 | — | |
| MQTReduction Ratio=88%, Vision Tokens=162026.02 | 61.9 | — | |
| FMVR-LLaVA#Vision Tokens=452026.03 | 61.9 | — | |
| DivPrune#Vision Tokens=1282026.03 | 61.5 | — |