Visual Question Answering on POPE v2
100AccuracyUpper Bound
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Upper BoundAvg. Token=5762026.04 | 100 | 302.4 | — | 403.1 | — | 9.6 | — | |
| RCPAvg. Token=1922026.04 | 99.06 | 100.8 | 66.7 | 263.7 | 34.6 | 3.52 | 63.3 | |
| RCPAvg. Token=1282026.04 | 96.73 | 67.2 | 77.8 | 196 | 51.4 | 2.4 | 75 | |
| RCPAvg. Token=642026.04 | 95 | 33.6 | 88.9 | 169.6 | 57.9 | 1.37 | 85.7 | |
| Qwen2-VLParameters=2B2025.12 | 91.3 | — | — | — | — | — | — | |
| SmolVLMParameters=2B2025.12 | 88.8 | — | — | — | — | — | — | |
| FastVAvg. Token=1922026.04 | 88.71 | 100.8 | 66.7 | 230.1 | 42.9 | 2 | 79.2 | |
| Qwen2-VLParameters=7B2025.12 | 87 | — | — | — | — | — | — | |
| VL-JEPASFTParameters=1.6B2025.12 | 86.3 | — | — | — | — | — | — | |
| SmolVLMParameters=500M2025.12 | 83.8 | — | — | — | — | — | — | |
| Qwen2-VLParameters=72B2025.12 | 79.4 | — | — | — | — | — | — | |
| InternVL2Parameters=26B2025.12 | 76.1 | — | — | — | — | — | — | |
| InternVL2Parameters=8B2025.12 | 74.5 | — | — | — | — | — | — | |
| LLaVA-1.5Parameters=13B2025.12 | 72.7 | — | — | — | — | — | — | |
| SmolVLMParameters=256M2025.12 | 62.3 | — | — | — | — | — | — |