Harmful behavior detection on Constructed in-domain dataset (test) benchmark leaderboard · SOTA2 Research

Harmful behavior detection on Constructed in-domain dataset (test)

99.2AccuracyWebAgentGuard-8B

Evaluation Results

Method	Links
WebAgentGuard-8B2026.04		99.2	98.4	100	99.19
WebAgentGuard-4B2026.04		98.2	96.8	99.59	98.17
GPT-4.12026.04		74.9	49.8	100	66.48
GPT-4o2026.04		70	40.2	100	57.34
GPT-4o-Mini2026.04		62.5	32	82.05	46.04
Qwen3-VL-Instruct-4B2026.04		58.2	17.33	100	29.55
Prompt-Guard-1-86M2026.04		56.2	93.4	53.56	68.08
Qwen3-VL-Instruct-8B2026.04		53.2	6.4	100	12.03
Llama-3.2-Vision-Instruct-11B2026.04		51.9	10.08	73.13	17.72
Qwen2.5-VL-Instruct-7B2026.04		50.9	3.51	73.91	6.7
Prompt-Guard-2-86M2026.04		50.7	2.6	68.42	5.01
GuardReasoner-VL-7B2026.04		50	0.2	50	3.98
Llama-Guard3-Vision-11B2026.04		49.6	1.4	38.89	2.7