Multimodal Evaluation on SEED-Bench (SEED-Bench Metric)
66.8SEED-Bench ScoreLLaVA-1.5-7B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LLaVA-1.5-7BReq. Inst.=100%, Sel. Inst.=665K2025.03 | 66.8 | 100 | |
| BaseBackbone=LLaVA-1.5-7B2026.03 | 66.1 | — | |
| Qwen-VL-ChatReq. Inst.=100%, Sel. Inst.=50M2025.03 | 65.4 | — | |
| FastVEnergy Threshold (τ)=99.8%, Adaptive Budget (+E)=false, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=2972026.03 | 65.4 | 98.9 | |
| FastV+EEnergy Threshold (τ)=99.8%, Adaptive Budget (+E)=true, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=2972026.03 | 65.2 | 98.6 | |
| VisionZipEnergy Threshold (τ)=99.8%, Adaptive Budget (+E)=false, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=2972026.03 | 65.2 | 98.6 | |
| VisionZip+EEnergy Threshold (τ)=99.8%, Adaptive Budget (+E)=true, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=2972026.03 | 65.1 | 98.5 | |
| PDropEnergy Threshold (τ)=99.8%, Adaptive Budget (+E)=false, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=2972026.03 | 64.5 | 97.6 | |
| PDrop+EEnergy Threshold (τ)=99.8%, Adaptive Budget (+E)=true, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=2972026.03 | 64.3 | 97.3 | |
| PreSelReq. Inst.=15%, Sel. Inst.=93K2025.03 | 63.8 | 97.9 | |
| VisionZipEnergy Threshold (τ)=99.0%, Adaptive Budget (+E)=false, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=1612026.03 | 63 | 95.3 | |
| VisionZip+EEnergy Threshold (τ)=99.0%, Adaptive Budget (+E)=true, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=1612026.03 | 63 | 95.3 | |
| TypiClustReq. Inst.=15%, Sel. Inst.=93K2025.03 | 62.8 | 96.8 | |
| FastVEnergy Threshold (τ)=99.0%, Adaptive Budget (+E)=false, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=1612026.03 | 62.5 | 94.6 | |
| RandomReq. Inst.=15%, Sel. Inst.=93K2025.03 | 62.4 | 95.7 | |
| FastV+EEnergy Threshold (τ)=99.0%, Adaptive Budget (+E)=true, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=1612026.03 | 62.3 | 94.3 | |
| COINCIDEReq. Inst.=100%, Sel. Inst.=93K2025.03 | 62 | 95.5 | |
| EL2NReq. Inst.=100%, Sel. Inst.=93K2025.03 | 61.8 | 93 | |
| PDrop+EEnergy Threshold (τ)=99.0%, Adaptive Budget (+E)=true, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=1612026.03 | 60.1 | 90.9 | |
| PDropEnergy Threshold (τ)=99.0%, Adaptive Budget (+E)=false, Backbone=LLaVA-1.5-7B, Mean Effective Token Count=1612026.03 | 60 | 90.8 | |
| IFDReq. Inst.=100%, Sel. Inst.=93K2025.03 | 59.4 | 91.8 | |
| InstructBLIP-7BReq. Inst.=100%, Sel. Inst.=1.2M2025.03 | 58.8 | — | |
| ShikraReq. Inst.=100%, Sel. Inst.=5.5M2025.03 | 58.8 | — | |
| PerplexityReq. Inst.=100%, Sel. Inst.=93K2025.03 | 58.3 | 91.9 | |
| Self-FilterReq. Inst.=100%, Sel. Inst.=93K2025.03 | 56.3 | 88.8 | |
| CLIP-ScoreReq. Inst.=100%, Sel. Inst.=93K2025.03 | 56.2 | 90.3 | |
| IDEFICS-80BReq. Inst.=100%, Sel. Inst.=1M2025.03 | 53.2 | — | |
| BLIP-2Req. Inst.=100%2025.03 | 49.7 | — |