Fine-grained multimodal reasoning on FGExpertBench benchmark leaderboard · SOTA2 Research

Fine-grained multimodal reasoning on FGExpertBench

74.88Object AccuracyKFRA

Evaluation Results

Method	Links
KFRA2026.03		74.88	74.49	77.55	71.33	75.33	75.29	74.81
Doubao-1.5-Vision-Pro2026.03		71.92	66.32	73.47	67.33	71.33	68.47	69.81
KFRA2026.03		69.54	72.45	83.67	65.33	70.67	69.41	71.85
Gemini-2.5-Flash2026.03		68.96	69.39	71.43	69.33	68.67	72.12	69.98
KFRA2026.03		68.47	64.29	75.51	68	64.67	65.76	67.78
GPT-4o2026.03		66.5	65.31	67.35	61.33	68.46	61.19	65.03
GLM-4.5V-12B2026.03		65.52	64.29	67.34	64.67	65.33	65.29	65.41
Qwen3-VL-A22B2026.03		64.53	61.22	73.47	66.67	67.33	66.24	66.58
GLM-4.1V-9B2026.03		62.56	63.27	57.14	64.67	59.33	58.71	60.95
Qwen-VL-Plus2026.03		60.1	57.14	75.51	66	56	61.53	62.71
Qwen2.5-VL-72B2026.03		57.64	51.02	69.39	69.33	66.67	63.65	62.95
Qwen3-VL-8B2026.03		51.72	63.27	75.51	53.33	56.67	56.71	59.54
GLM-4V-Plus2026.03		47.29	45.92	63.27	64.67	58	48.35	54.58
Gemma3-it-4B2026.03		45.81	51.02	55.1	51.33	45.33	54	50.43
InternVL3-8B2026.03		42.36	52.04	61.22	58	55.33	52.24	53.53
InternVL3.5-8B2026.03		39.41	47.96	65.31	61.33	54.66	51.18	53.31
Phi-3.5-4B2026.03		36.45	46.94	51.02	47.33	56	45.76	47.47
Qwen2.5-VL-7B2026.03		33.5	42.86	63.27	46.67	34	41.53	48.64
VSA2026.03		31.68	44.9	51.02	31.33	38.67	37.81	39.24
LLaVA-Next-8B2026.03		31.53	51.02	57.14	44	36.67	42.24	43.77
LLaVA-OneVision-7B2026.03		31.03	36.73	48.98	40.67	34.67	44.27	39.39
MMSearch2026.03		27.09	39.8	38.78	26	48	41.65	36.81