List your company

Object Hallucination Detection on POPE (test)

89.4Final Performance ScoreMAI
87.00887.62988.2588.871Apr 6, 2026

Evaluation results

MethodLinks
MAIBackbone=Qwen3-VL-2B/8B2026.0489.48,50746447
Full RTBackbone=Qwen3-VL-8B+2B-R2026.0489.2———
2B(S) BaselineBackbone=Qwen3-VL-2B, Mode=Standard2026.0488.6———
2B(R) BaselineBackbone=Qwen3-VL-2B, Mode=Reasoning2026.0488.4———
8B(S) BaselineBackbone=Qwen3-VL-8B, Mode=Standard2026.0487.1———