Visual Misinterpretation Hallucination Open-ended (test) benchmark leaderboard · SOTA2 Research

Visual Misinterpretation Hallucination Open-ended (test)

35.8CheXbert ScoreMedQwen

Evaluation Results

Method	Links
MedQwen2026.04		35.8	13.78	49.77	37.24	6.21
CheXagent2026.04		30.32	12.35	43.18	33.93	6.88
XrayGPT2026.04		25.63	12.88	44.45	30.87	12.84
LLM-CXR2026.04		24.34	7.57	38.53	29.85	9.18
RadFM2026.04		23.74	6.69	37.04	24.66	6.89
GPT-4o2026.04		21.71	10.28	45.39	33.73	11.99
LLaVA-Med2026.04		19.72	7.31	39.86	25.17	20.85
LLaVA-Med-1.52026.04		18.44	4.96	39.47	13.27	19.74
MiniGPT-42026.04		17.71	7.18	39.9	10.54	18.02
Med-Flamingo2026.04		17.5	5.83	35.87	17.52	23.96
LLaVA-NeXT2026.04		16.31	4.41	39.93	10.88	16.08
LLaVA-NeXT2026.04		14.76	5.34	38.59	6.38	14.82