TRACE: Trajectory Reasoning through Adaptive Cross-Step Evidence Aggregation for LLM Agents

In three linesTRACE is a monitoring framework for long-horizon LLM agent trajectories. It uses a TIJ (Triage-Inspect-Judge) loop to identify high-signal regions, inspect actions while accumulating evidence, and synthesize trajectory-level verdicts. Evaluated on SHADE-Arena: F1=0.713, recall=0.844.

Summary generated by Claude — human-verified