Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces OAT, an unsupervised failure attribution model for LLM-based agentic systems that identifies error steps in failure trajectories without requiring step-level supervision. By leveraging neural controlled differential equations to model the dynamics of successful trajectories, OAT assigns anomaly scores to steps in failure trajectories based on their deviation from learned patterns. The method significantly outperforms existing prompting-based approaches, achieving 200–5000 times faster processing while improving F1 scores by +20% and +7% on in-domain and out-of-distribution datasets, respectively.
OAT achieves up to 5000 times faster failure attribution for LLM agents without the need for costly step-level supervision.
Failure attribution for LLM-based agentic systems, i.e., identifying which steps in a failure trajectory caused the task to fail, is critical for debugging and improving these systems. Existing approaches either rely on prompting-based pipelines, which are computationally expensive, or require post-training on failure trajectories with step-level error annotations, which are costly to collect and difficult to scale. We argue that a practical failure attribution model should be lightweight and trainable without step-level supervision on failure data. To this end, we address unsupervised failure attribution, i.e., training exclusively on successful trajectories and identifying error steps at inference time given a failure trajectory. We propose OAT, which casts this problem as one-class learning with neural controlled differential equations, modeling the dynamical pattern of successful trajectories in latent space. At inference time, each step in a failure trajectory is assigned an anomaly score based on its deviation from the dynamics learned on successful trajectories, which is then used to form a set of error steps. With training on only 100 successful trajectories, experiments show that OAT is 200--5000 $\times$ faster than prompting-based baselines, and, at the same time, consistently outperforms them in both in-domain and out-of-distribution datasets with +20% and +7% F1 scores, respectively, demonstrating that OAT is a promising and efficient direction for diagnosing agentic system failures.