Search papers, labs, and topics across Lattice.
This paper introduces a method to transform long unstructured traces of LLM-based agents into a compact finite-state machine (FSM), enabling improved next-step and failure predictions. By analyzing twelve public datasets, the FSMs demonstrated high fidelity in replaying held-out data and revealed a consistent behavioral topology across different splits. The findings suggest that the deployment environment significantly influences agent behavior, providing a model-agnostic framework for enhancing safety auditing and monitoring in AI systems.
Compact finite-state machines can predict LLM agent failures with up to 94% accuracy, revealing that deployment harnesses shape behavior more than the models themselves.
LLM-based agents execute multi-step tasks, but their behavioral structure remains opaque: long unstructured traces resist the safety auditing and runtime monitoring that deployment requires. Existing approaches operate per-trace or success-only, so they miss the cross-run topology that links next-step and failure prediction. To recover that shared structure, we collapse an entire trace corpus into a single, compact finite-state machine (FSM) that serves as a structural substrate for the otherwise unpredictable behavior of LLM agents. Across twelve public datasets, the FSMs are compact (7-43 states), replay held-out data at>=0.997 fitness with near-identical topology across splits, and build in milliseconds. This substrate addresses both prediction goals. For next-step prediction, FSM-state context outperforms Agent Workflow Memory on every ground-truth-matched dataset. For failure prediction, per-state behavioral features reach held-out AUROC up to 0.94, and an online monitor ranks failing runs above passing ones from a partial trace, triggering early stopping well before completion. Behavioral topology thus appears shaped more by the deployment harness than by the LLM, providing a model-agnostic structural primitive for safety auditing and runtime monitoring.
