List your company

Vision-Language Hallucination Assessment on MM-HallBench

3.64Average ScoreGreedy
3.30723.39363.483.5664Apr 25, 2026

Evaluation results

MethodLinks
GreedyBackbone=Qwen2.5-VL (3B)2026.043.6434
Top-p (nucleus)Backbone=Qwen2.5-VL (3B)2026.043.5836
η-SamplingBackbone=Qwen2.5-VL (3B)2026.043.5738
Beam SearchBackbone=Qwen2.5-VL (3B)2026.043.5434
Temp’ OnlyBackbone=Qwen2.5-VL (3B)2026.043.4836
Top-kBackbone=Qwen2.5-VL (3B)2026.043.3842
ε-SamplingBackbone=Qwen2.5-VL (3B)2026.043.3635
Min-pBackbone=Qwen2.5-VL (3B)2026.043.3243