StepFinder: A Temporal Semantic Framework for Failure Attribution in Multi-Agent Systems

2026-06-02

AI Summary

This work addresses the challenge of step-level root cause localization in multi-agent systems, where single-step execution errors can trigger cascading failures. Existing large language model (LLM)-based attribution methods suffer from high latency, substantial computational cost, and susceptibility to redundant log noise due to their reliance on full-trajectory reasoning. To overcome these limitations, we propose StepFinder, a novel framework that leverages LLMs solely for semantic feature encoding of logs—eschewing end-to-end reasoning—and integrates lightweight temporal modeling with an attention mechanism to capture inter-step evolution and dependencies. StepFinder further refines error scoring through multi-scale discrepancy analysis and positional bias correction. Evaluated on the Who&When benchmark, StepFinder substantially outperforms current LLM-based approaches, reducing inference time by 79% while eliminating text generation overhead and achieving highly accurate, efficient root cause identification at the step level.

Abstract

LLM-based multi-agent systems exhibit remarkable collaborative capabilities in complex multi-step tasks. However, these systems are highly sensitive to single-step execution errors that can propagate through agent interactions and lead to cascading failures. To understand the causes of failure and improve system reliability, failure attribution has been introduced as a task that aims to automatically identify the root cause step responsible for a failure. Existing failure attribution methods mainly rely on LLMs to reason over original execution trajectories, which not only incur high inference costs and latency, but also suffer from interference caused by redundant and noisy execution logs, causing LLMs to struggle in accurately identifying the true root cause step. To address this, we propose StepFinder, a lightweight failure attribution framework. We use LLMs solely during the feature construction phase to encode execution logs into temporal semantic sequences. Subsequently, a parameter-efficient combination of temporal modeling and attention modules is applied to capture the sequential evolution and cross-step dependencies of the trajectories. Finally, the step-level error score is refined through multi-scale differences and position bias, enabling precise root cause identification. Experimental results on the Who&When benchmark demonstrate that StepFinder outperforms LLM-based methods in step-level failure attribution while achieving substantially higher inference efficiency, reducing inference time by 79% compared with the fastest LLM-based method, with no text generation overhead. Our code is available at https://github.com/taiyu-zhu/StepFinder.

Problem

• failure attribution
• multi-agent systems
• root cause identification
• execution errors
• LLM-based reasoning

Innovation

• failure attribution
• temporal semantic framework
• multi-agent systems
• parameter-efficient modeling
• root cause identification