Our new X account is live! Follow @wizwand_team for updates

Step-level tool invocation safety detection on AgentHarm Traj

84.81Accuracy

TS-Guard

Updated 4d ago

Evaluation Results

Method	Links
TS-Guard 2026.01		84.81	90.16	96.95
Llama-Guard-3-8B 2026.01		81.53	86.35	81.33
Qwen3Guard-8B-Gen 2026.01		80.57	86.27	84.95
TS-Guard 2026.01		79.93	75.93	75.99
gpt-4o 2026.01		79.48	63.94	63.33
Qwen3-8B 2026.01		76.47	47.24	36.67
GPT-4o 2026.01		75.23	84.8	96.19
Qwen2.5-7B-IT 2026.01		74.28	29.32	18.57
ShieldAgent-THU 2026.01		71.13	82.63	95.62
Qwen3Guard-8B-Gen 2026.01		70.72	47.8	46.67
Qwen2.5-7B-IT 2026.01		67.99	80.17	90.09
TS-Guard 2026.01		64.84	55.92	77.62
Safiron 2026.01		64.02	36.32	35.71
gpt-4o 2026.01		58.14	53.63	54.53
Llama-Guard-3-8B 2026.01		56.77	53.11	85.24
Qwen3-8B 2026.01		54.46	58.94	45.52
Safiron 2026.01		46.1	45.88	31.81
Qwen3-8B 2026.01		45.14	44.95	47.9
Qwen2.5-7B-IT 2026.01		45	34.85	38.22
ShieldAgent-THU 2026.01		33.24	45.78	98.09