List your company

Tool-use on AgentHarm (public test)

50TSRbaseline
-211.52538.5Jun 14, 2026

Evaluation results

MethodLinks
baselineAccess Control=LLM-AC, Core LLM=GPT-4o2026.06503.4
direct queryingAccess Control=None, Core LLM=GPT-4o2026.0638.6—
baselineAccess Control=LLM-AC, Core LLM=GPT-5.12026.0632.421
baselineAccess Control=ShieldAgent, Core LLM=Gemini 2.5 Flash2026.0628.921.6
baselineAccess Control=LLM-AC, Core LLM=Average2026.0627.58.1
direct queryingAccess Control=None, Core LLM=Average2026.0625—
FragFuseAccess Control=LLM-AC, Core LLM=GPT-4o2026.062545.5
direct queryingAccess Control=None, Core LLM=Gemini 2.5 Flash2026.0622.7—
FragFuseAccess Control=ShieldAgent, Core LLM=Gemini 2.5 Flash2026.0619.385.2
FragFuseAccess Control=ShieldAgent, Core LLM=GPT-4o2026.0617.282.4
direct queryingAccess Control=None, Core LLM=GPT-5.12026.0613.6—
FragFuseAccess Control=ShieldAgent, Core LLM=Average2026.0613.188.3
baselineAccess Control=ShieldAgent, Core LLM=Average2026.061318.7
FragFuseAccess Control=LLM-AC, Core LLM=Average2026.0610.980.5
baselineAccess Control=ShieldAgent, Core LLM=GPT-5.12026.0610.233.5
FragFuseAccess Control=LLM-AC, Core LLM=GPT-5.12026.067.697.2
FragFuseAccess Control=ShieldAgent, Core LLM=GPT-5.12026.062.997.2
baselineAccess Control=LLM-AC, Core LLM=Gemini 2.5 Flash2026.0600
FragFuseAccess Control=LLM-AC, Core LLM=Gemini 2.5 Flash2026.06098.9
baselineAccess Control=ShieldAgent, Core LLM=GPT-4o2026.0601.1