List your company

Trajectory Auditing on AFTRAJ-2K (D_safe D_unsafe)

2.37False Alarm Rate (FAR)AgentForesight-7B
-1.440424.27985075.7202May 9, 2026

Evaluation results

MethodLinks
AgentForesight-7BModel Category=Ours2026.052.3759.51
DeepSeek-V4-ProModel Category=Proprietary LLMs2026.0543.253.99
Qwen2.5-7B-InstructModel Category=Open-Source LLMs2026.0546.1536.2
Qwen3-8BModel Category=Open-Source LLMs2026.0556.838.04
DeepSeek-V4-FlashModel Category=Proprietary LLMs2026.0559.7647.24
Gemini-3-FlashModel Category=Proprietary LLMs2026.0567.8638.04
Claude-Haiku-4.5Model Category=Proprietary LLMs2026.0568.6433.13
GPT-4.1Model Category=Proprietary LLMs2026.0585.838.04
Llama3.2-3BModel Category=Open-Source LLMs2026.0590.5320.86
Gemma3-4BModel Category=Open-Source LLMs2026.0597.6310.43