Grounded Visual Question Answering on Visual7W (test)
92.3AccuracyGPT-4o
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GPT-4ozero-shot=true, tool-assisted=true2026.04 | 92.3 | 73.3 | 90.8 | 45.5 | 51 | 66.9 | |
| GPT-4ochain-of-thought=true2026.04 | 91.8 | 73.8 | 90.3 | 33.4 | 24 | 57 | |
| CogVLM-GroundingType=Generalist2023.11 | 91.05 | — | — | — | — | — | |
| Geminichain-of-thought=true2026.04 | 89.9 | 71.9 | 89.4 | 38.2 | 40 | 61.4 | |
| AutoVQA-G2026.04 | 89.6 | 73.5 | 81.9 | 63.4 | 72 | 74.7 | |
| Geminizero-shot=true2026.04 | 89.4 | 70.8 | 87.7 | 39 | 44 | 62.1 | |
| Human Annotation2026.04 | 89 | 65.1 | 86.5 | 51.7 | 56 | 67 | |
| Shikra-13BType=Generalist2023.11 | 85.33 | — | — | — | — | — |