Object Hallucination Detection on POPE (test)
89.4Final Performance ScoreMAI
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MAIBackbone=Qwen3-VL-2B/8B2026.04 | 89.4 | 8,507 | 46 | 447 | |
| Full RTBackbone=Qwen3-VL-8B+2B-R2026.04 | 89.2 | — | — | — | |
| 2B(S) BaselineBackbone=Qwen3-VL-2B, Mode=Standard2026.04 | 88.6 | — | — | — | |
| 2B(R) BaselineBackbone=Qwen3-VL-2B, Mode=Reasoning2026.04 | 88.4 | — | — | — | |
| 8B(S) BaselineBackbone=Qwen3-VL-8B, Mode=Standard2026.04 | 87.1 | — | — | — |