Visual Instruction Following Evaluation on LLaVA-Bench
4.38AccuracyDeFG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DeFGBackbone=InstructBLIP, Reference Baseline=Regular2025.12 | 4.38 | 4.79 | |
| CoFi-DecBackbone=LLaVA-1.5, Reference Baseline=Regular2025.12 | 4.33 | 4.57 | |
| CoFi-DecBackbone=InstructBLIP, Reference Baseline=Regular2025.12 | 4.33 | 4.86 | |
| DeFGBackbone=LLaVA-1.5, Reference Baseline=Regular2025.12 | 4.29 | 4.54 | |
| CoFi-DecBackbone=InstructBLIP, Reference Baseline=M3ID2025.12 | 4.27 | 4.51 | |
| CoFi-DecBackbone=InstructBLIP, Reference Baseline=VCD2025.12 | 4.26 | 4.63 | |
| DeFGBackbone=InstructBLIP, Reference Baseline=VCD2025.12 | 4.17 | 4.58 | |
| CoFi-DecBackbone=LLaVA-1.5, Reference Baseline=M3ID2025.12 | 4.13 | 4.18 | |
| CoFi-DecBackbone=LLaVA-1.5, Reference Baseline=VCD2025.12 | 4.12 | 4.31 | |
| DeFGBackbone=InstructBLIP, Reference Baseline=M3ID2025.12 | 4.08 | 4.5 | |
| DeFGBackbone=LLaVA-1.5, Reference Baseline=VCD2025.12 | 4.04 | 4.38 | |
| DeFGBackbone=LLaVA-1.5, Reference Baseline=M3ID2025.12 | 4.04 | 4.29 | |
| M3IDBackbone=InstructBLIP2025.12 | 4 | 4.33 | |
| M3IDBackbone=LLaVA-1.52025.12 | 3.88 | 4.08 | |
| VCDBackbone=InstructBLIP2025.12 | 3.71 | 4.21 | |
| VCDBackbone=LLaVA-1.52025.12 | 3.62 | 3.83 | |
| RegularBackbone=InstructBLIP2025.12 | 3.42 | 3.96 | |
| RegularBackbone=LLaVA-1.52025.12 | 2.88 | 3.29 |