Anomaly Reasoning on MVTec-COT (test) benchmark leaderboard · SOTA2 Research

Anomaly Reasoning on MVTec-COT (test)

27.1ROUGE-LQwen2.5-VL+CGRO*

Evaluation Results

Method	Links
Qwen2.5-VL+CGRO*2026.03		27.1	74.7
Qwen2.5-VL+R1*2026.03		26.3	73.8
Qwen2.5-VL+CGRO*2026.03		25.6	72.4
Qwen2.5-VL+R1*2026.03		24.2	71.7
LLaVA-OneVision-OV2026.03		23.7	70.5
InternVL-32026.03		22.2	70.1
MiniCPM-V-2.62026.03		21.7	68.3
Qwen2.5-VL2026.03		21.1	67.5
EIAD2026.03		21.1	65.7
GPT-4.12026.03		20.8	69.9
Qwen2.5-VL2026.03		20.8	65.5
Qwen2.5-VL2026.03		20	66.1
GPT-4o2026.03		19.9	60.4
InternVL-2.52026.03		19.7	57.5
Qwen3VL-thinking2026.03		19.3	61.9
GPT-4o-mini2026.03		19	65.2
LLaVA-1.62026.03		18.6	54.2
Gemini-2.5-Flash2026.03		17.6	68.1
Qwen3VL-thinking-A3B2026.03		16.9	58.9
Qwen2.5-VL2026.03		16.8	49
GPT-5-nano2026.03		14	50
AnomalyGPT2026.03		11.9	36.7
LLaVA-OneVision-SI2026.03		10.5	39.1
Triad2026.03		8.6	35.9