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

Step-level tool invocation safety detection on AgentDojo-Traj

91.72Accuracy

TS-Guard

Updated 4d ago

Evaluation Results

Method	Links
TS-Guard 2026.01		91.72	86.18	89.49
TS-Guard 2026.01		85.81	68.26	52.84
Qwen3-8B 2026.01		85.47	77.1	84.9
gpt-4o 2026.01		84.92	79.14	99.15
TS-Guard 2026.01		81.15	54.11	50.48
Qwen3-8B 2026.01		79.65	72.32	92.31
Qwen3-8B 2026.01		77.52	53.67	53.15
Llama-Guard-3-8B 2026.01		74.75	33.33	21.87
Llama-Guard-3-8B 2026.01		74.75	33.33	21.87
Qwen2.5-7B-IT 2026.01		74.73	50.95	45.58
Qwen3Guard-8B-Gen 2026.01		71.15	0	0
Safiron 2026.01		70.98	52.8	56.25
Safiron 2026.01		70.98	52.8	56.25
Qwen3Guard-8B-Gen 2026.01		70.57	3.23	1.7
ShieldAgent-THU 2026.01		60.9	58.91	97.16
ShieldAgent-THU 2026.01		60.9	58.91	97.16
GPT-4o 2026.01		55.74	56.59	100
gpt-4o 2026.01		55.49	44.66	45.64
Qwen2.5-7B-IT 2026.01		43.97	50.47	99.14
Qwen2.5-7B-IT 2026.01		28.55	28.82	22.41