Share your thoughts, 1 month free Claude Pro on usSee more

Cell-Level Visual Question Answering on PBSBench (In-domain)

77True/False Accuracy

GPT-4o

Updated 1mo ago

Evaluation Results

Method	Links
GPT-4o 2026.04		77	90	23	48	9	14	65
InternVL-8B 2026.04		73	91	14	38	13	18	63
PBS-VL 2026.04		73	96	47	54	36	40	72
Gemini-1.5-pro 2026.04		70	88	3	41	6	10	60
Qwen2-VL-7B 2026.04		67	81	16	27	10	14	62
PathGen-LLaVA 2026.04		65	78	15	31	14	19	63
Qwen2.5-VL-7B 2026.04		63	76	3	17	4	7	59
DeepSeek-VL-7B 2026.04		63	51	0	15	2	4	52
Quilt-LLaVA 2026.04		61	36	9	16	9	13	59
Claude-3.5-Sonnet 2026.04		59	72	2	54	4	8	50
Idefics3-8B 2026.04		59	35	11	17	14	17	56
LLaVA-1.6-Mistral-7B 2026.04		59	52	6	10	5	8	56
LLaVA-1.6-Vicuna-7B 2026.04		58	46	7	10	4	7	54
LLaVA-Med 2026.04		54	29	0	10	11	15	56
BLIP2-OPT-6.7B 2026.04		51	20	1	21	9	11	41
Med-flamingo 2026.04		45	27	0	6	7	9	43
PA-LLaVA 2026.04		39	24	0	0	1	2	5