Multimodal Evaluation on MME (Total)
2,631.7Total ScoreQwen3-VL-235B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-VL-235BToken Retention Ratio=100%2026.05 | 2,631.7 | |
| VisionZipToken Retention Ratio=60%2026.05 | 2,588.9 | |
| F3AToken Retention Ratio=60%2026.05 | 2,580.3 | |
| DivPruneToken Retention Ratio=60%2026.05 | 2,562.4 | |
| CDPrunerToken Retention Ratio=60%2026.05 | 2,548.2 | |
| FastVToken Retention Ratio=60%2026.05 | 2,511.6 | |
| VisionZipToken Retention Ratio=40%2026.05 | 2,440.1 | |
| F3AToken Retention Ratio=40%2026.05 | 2,439.7 | |
| DivPruneToken Retention Ratio=40%2026.05 | 2,428.4 | |
| CDPrunerToken Retention Ratio=40%2026.05 | 2,412.6 | |
| FastVToken Retention Ratio=40%2026.05 | 2,366.5 | |
| F3AToken Retention Ratio=20%2026.05 | 2,366.1 | |
| DivPruneToken Retention Ratio=20%2026.05 | 2,291.6 | |
| CDPrunerToken Retention Ratio=20%2026.05 | 2,268.4 | |
| VisionZipToken Retention Ratio=20%2026.05 | 2,240.5 | |
| VanillaBackbone=InternVL-2.5-8B, Token Reduction Rate=100%2026.04 | 2,240 | |
| FastVToken Retention Ratio=20%2026.05 | 2,104.7 | |
| DASHBackbone=InternVL-2.5-8B, Token Reduction Rate=25%2026.04 | 2,098 | |
| DivPruneBackbone=InternVL-2.5-8B, Token Reduction Rate=25%2026.04 | 2,097 | |
| DARTBackbone=InternVL-2.5-8B, Token Reduction Rate=25%2026.04 | 2,086 | |
| FastVBackbone=InternVL-2.5-8B, Token Reduction Rate=25%2026.04 | 2,073 | |
| VanillaRetained Tokens=576, Backbone=LLaVA-v1.5-7B2026.05 | 1,862 | |
| LRCPRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 1,823 | |
| ApETRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 1,808 | |
| LRCPRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 1,807 | |
| ApETRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 1,801 | |
| V2DropRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 1,796 | |
| VisionZipRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 1,778 | |
| PDropRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 1,766 | |
| VisionZipRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 1,759 | |
| SparseVLMRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 1,721 | |
| LRCPRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 1,721 | |
| ApETRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 1,714 | |
| V2DropRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 1,712 | |
| SparseVLMRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 1,696 | |
| VisionZipRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 1,687 | |
| PDropRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 1,644 | |
| FastVRetained Tokens=192, Backbone=LLaVA-v1.5-7B2026.05 | 1,612 | |
| SparseVLMRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 1,505 | |
| COINCIDETraining Data Percentage=20%, Data Selection Constraint=Full Data Used before Selection2026.04 | 1,495.6 | |
| FastVRetained Tokens=128, Backbone=LLaVA-v1.5-7B2026.05 | 1,490 | |
| ICONSTraining Data Percentage=20%, Data Selection Constraint=Full Data Used before Selection2026.04 | 1,485.7 | |
| FullTraining Data Percentage=100%, Data Selection Constraint=None (Oracle)2026.04 | 1,476.9 | |
| V2DropRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 1,470 | |
| DOSETraining Data Percentage=20%, Data Selection Constraint=Partial Data Used before Selection2026.04 | 1,462.2 | |
| EL2NTraining Data Percentage=20%, Data Selection Constraint=Partial Data Used before Selection2026.04 | 1,439.5 | |
| D2-PruningTraining Data Percentage=20%, Data Selection Constraint=Partial Data Used before Selection2026.04 | 1,391.2 | |
| RandomTraining Data Percentage=20%, Data Selection Constraint=Partial Data Used before Selection2026.04 | 1,389 | |
| SemDeDupTraining Data Percentage=20%, Data Selection Constraint=Partial Data Used before Selection2026.04 | 1,376.9 | |
| PerplexityTraining Data Percentage=20%, Data Selection Constraint=Partial Data Used before Selection2026.04 | 1,341.4 | |
| Self-SupTraining Data Percentage=20%, Data Selection Constraint=Partial Data Used before Selection2026.04 | 1,335.9 | |
| CLIP-ScoreTraining Data Percentage=20%, Data Selection Constraint=Partial Data Used before Selection2026.04 | 1,331.6 | |
| Self-FilterTraining Data Percentage=20%, Data Selection Constraint=Partial Data Used before Selection2026.04 | 1,306.2 | |
| FastVRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 1,256 | |
| PDropRetained Tokens=64, Backbone=LLaVA-v1.5-7B2026.05 | 1,092 | |
| InternVL2-8B + ACTArchitecture=InternVL2-8B, ACT=Enabled2026.04 | 711.7 | |
| InternVL2.5-8B + ACTArchitecture=InternVL2.5-8B, ACT=Enabled2026.04 | 710 | |
| InternVL2.5-8BArchitecture=InternVL2.5-8B, ACT=Disabled2026.04 | 705 | |
| Qwen2.5VL-7B + ACTArchitecture=Qwen2.5VL-7B, ACT=Enabled2026.04 | 705 | |
| InternVL2-8BArchitecture=InternVL2-8B, ACT=Disabled2026.04 | 701.7 | |
| Qwen2.5VL-7BArchitecture=Qwen2.5VL-7B, ACT=Disabled2026.04 | 700 | |
| Qwen2VL-7B + ACTArchitecture=Qwen2VL-7B, ACT=Enabled2026.04 | 693.3 | |
| Qwen2VL-7BArchitecture=Qwen2VL-7B, ACT=Disabled2026.04 | 688.3 | |
| LLaVA-1.5-7B + ACTArchitecture=LLaVA-1.5-7B, ACT=Enabled2026.04 | 668.3 | |
| LLaVA-1.5-13B + ACTArchitecture=LLaVA-1.5-13B, ACT=Enabled2026.04 | 645 | |
| LLaVA-1.5-7BArchitecture=LLaVA-1.5-7B, ACT=Disabled2026.04 | 638.3 | |
| LLaVA-1.5-13BArchitecture=LLaVA-1.5-13B, ACT=Disabled2026.04 | 636.7 |