Visual Metonymy Interpretation on ViMET Overall 1.0 (Full)
86.9AccuracyHuman Performance
Evaluation Results
| Method | Links | |
|---|---|---|
| Human Performance2026.01 | 86.9 | |
| InternVL3 78BModel Size=78B2026.01 | 65.9 | |
| Qwen2.5-VL 72BModel Size=72B2026.01 | 65.4 | |
| Gemini 2.5 Pro2026.01 | 65.2 | |
| Qwen2.5-VL 7BModel Size=7B2026.01 | 63.7 | |
| Llama 4 Scout2026.01 | 63.2 | |
| InternVL3 8BModel Size=8B2026.01 | 62.9 | |
| Llama 3.2 90BModel Size=90B2026.01 | 62.7 | |
| Gemini 2.5 Flash2026.01 | 62.6 | |
| Llama 3.2 11BModel Size=11B2026.01 | 61.2 |