Vision-Language Evaluation on MME full
2,418.5MME ScoreGIFT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GIFTModel=Qwen3-VL 8B2025.10 | 2,418.5 | — | — | — | |
| GreedyModel=Qwen3-VL 8B2025.10 | 2,403.5 | — | — | — | |
| GIFTModel=Qwen2-VL 7B2025.10 | 2,279.1 | — | — | — | |
| GreedyModel=Qwen2-VL 7B2025.10 | 2,278.9 | — | — | — | |
| GIFTModel=LLaVA-1.5 13B2025.10 | 1,815.9 | — | — | — | |
| VAFModel=LLaVA-1.5 13B2025.10 | 1,815.4 | — | — | — | |
| Rel-AttnModel=LLaVA-1.5 7B2025.10 | 1,811.3 | — | — | — | |
| GreedyModel=LLaVA-1.5 13B2025.10 | 1,807.5 | — | — | — | |
| VCDModel=LLaVA-1.5 13B2025.10 | 1,804 | — | — | — | |
| VAFModel=LLaVA-1.5 7B2025.10 | 1,787.6 | — | — | — | |
| VARModel=LLaVA-1.5 13B2025.10 | 1,782.5 | — | — | — | |
| VARModel=LLaVA-1.5 7B2025.10 | 1,780.8 | — | — | — | |
| Rel-AttnModel=LLaVA-1.5 13B2025.10 | 1,758.5 | — | — | — | |
| GreedyModel=LLaVA-1.5 7B2025.10 | 1,751.6 | — | — | — | |
| GIFTModel=LLaVA-1.5 7B2025.10 | 1,750.5 | — | — | — | |
| VCDModel=LLaVA-1.5 7B2025.10 | 1,742.3 | — | — | — | |
| AGLA2026.02 | — | 1,716.8 | 611.79 | 2,328.59 | |
| M3ID2026.02 | — | 1,715.73 | 611.79 | 2,327.52 | |
| Only2026.02 | — | 1,686.78 | 603.21 | 2,289.99 | |
| Regulardecoding=standard greedy decoding baseline2026.02 | — | 1,715.73 | 611.79 | 2,327.52 | |
| REVIS2026.02 | — | 1,717.44 | 627.86 | 2,345.3 | |
| VCD2026.02 | — | 1,665.65 | 593.21 | 2,258.86 | |
| VTI2026.02 | — | 1,492.39 | 591.43 | 2,083.82 |