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

Visual Reasoning on VLM2-Bench

95.06Mat

Human-Level

Updated 1mo ago

Evaluation Results

Method	Links
Human-Level 2026.02		95.06	98.11	96.02	94.23	91.29	97.08	92.87	91.17	100	95.16	0
ReMoT-4B-CoT 2026.02		57.23	54.14	87.56	73.06	78.5	92	76.67	66	46.75	70.21	-24.95
MiCo-7B-CoT 2026.02		57.14	67.12	81.94	56.67	58	65	57.5	62	44.25	61.06	-34.09
MM-Eureka-7B 2026.02		55.6	47.03	74.1	52.5	54	77.5	60	51	43.5	57.24	-37.91
Qwen3-VL-4B-CoT 2026.02		52.88	51.59	86.81	69.64	74.5	90	75	57	40.5	66.43	28.73
VLAA-Thinker-7B 2026.02		47.49	63.03	72.2	61.4	55	71	57.5	51	47.75	58.49	-36.67
Qwen2.5-VL-7B-CoT 2026.02		43.24	42.92	66.39	50.56	36	62.5	55.83	39	36.75	48.91	-46.24
InternVL2.5-8B 2026.02		41.24	26.53	72.22	67.65	40	85	66.67	52.25	50.25	55.41	-39.75
NoisyRollout-7B 2026.02		40.93	43.83	63.33	50.83	34.5	70.5	63.33	47	36.5	50.08	-45.08
ThinkLite-VL-7B 2026.02		40.45	46.58	75.56	62.5	49.5	77.5	62.5	51	36.5	55.79	-39.37
GPT-4o 2026.02		37.45	39.27	74.17	80.62	57.5	50	90.5	47	66.75	60.36	-34.8
Qwen2.5-VL-7B 2026.02		35.91	43.38	71.39	41.72	47.5	80	59.76	69	45	54.82	-40.34
InternVL2.5-26B 2026.02		30.5	30.59	43.33	51.48	52.5	59.5	59.67	61.25	45.25	45.59	-49.57
Chance-Level 2026.02		25	25	50	34.88	25	50	34.87	25	-	32.73	-61.44
LLaVA-Video-7B 2026.02		18.53	12.79	54.72	62.47	28.5	62	66.91	25	59	45.65	-49.51
Qwen2-VL-7B 2026.02		18.07	19.18	68.08	61.84	37.5	72	67.92	47	55.25	49.76	-45.4
mPLUG-Owl2-7B 2026.02		17.37	18.26	49.17	62.97	31	63	58.06	29	43	40.87	-54.31
LLaVA-OneVision 2026.02		16.6	13.7	47.22	56.17	27.5	62	46.67	37	47.25	39.35	-55.81
LongVQA-7B 2026.02		14.29	12.98	46.53	49.47	29	58	41.56	25	45	37.1	-58.06