Visual Question Answering on GQA (Score, Speed-up)
66.4ScoreBAGEL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BAGELSize=14B2026.04 | 66.4 | — | |
| Tuna-2Size=7B2026.04 | 65 | — | |
| TunaSize=7B2026.04 | 63.9 | — | |
| Tuna-RSize=7B2026.04 | 63.5 | — | |
| Show-o2Size=7B2026.04 | 63.1 | — | |
| OneCatSize=9B2026.04 | 63.1 | — | |
| X-OmniSize=7B2026.04 | 62.8 | — | |
| TokenFlow-XLSize=14B2026.04 | 62.5 | — | |
| LLaVA-1.5Size=7B2026.04 | 62 | — | |
| Janus-ProSize=7B2026.04 | 62 | — | |
| TarSize=7B2026.04 | 61.3 | — | |
| VILA-USize=7B2026.04 | 60.8 | — | |
| Qwen2.5-VLSize=7B2026.04 | 60.7 | — | |
| JanusFlowSize=1.3B2026.04 | 60.3 | — | |
| Emu3Size=8B2026.04 | 60.3 | — | |
| Ming-UniVisionSize=16B2026.04 | 59.4 | — | |
| HarmonSize=1.5B2026.04 | 58.9 | — | |
| Qwen-VL-ChatSize=7B2026.04 | 57.5 | — | |
| AHD (Ours)2026.04 | 50.58 | 16.4 | |
| MMaDA2026.04 | 50.52 | 1 |