Skip to content
Preprint

A Mechanistic Study of Automated Failure Attribution in LLM-based Agentic Systems

Oct 2026 · 0 citations · 28 references
Computer Science

TL;DR

Grafting is proposed, a simple manual intervention that exploits knowledge already encoded in the model to improve AFA performance without requiring ground-truth labels at inference time, and extensive evaluations demonstrate that grafting consistently outperforms direct-prompting approaches.

Abstract

Agentic systems have demonstrated strong capabilities on complex tasks by enabling multiple interacting large language model (LLM)-based agents to collaborate and execute subtasks in parallel. However, this collaborative structure introduces new challenges for reliably evaluating and diagnosing system failures, particularly failure attribution: tracing a system's execution to identify the agent step responsible for an incorrect final outcome. Because manually performing this task is time- and labor-intensive, recent work has turned to LLMs as judges to automate the process, referred to as automated failure attribution (AFA). These approaches assume that a sufficiently capable model can infer the failing step directly from an interaction transcript. In practice, however, direct prompting often performs poorly, which motivates follow-up work that explores various ways to improve AFA performance, including alternative prompting strategies and inference-time interventions. In this paper, we examine a more fundamental question: whether this weak behavioral performance reflects a genuine capability limitation. Using linear probes applied to internal activations, we uncover a significant \emph{knows-but-won't-say gap:} models encode sufficient information to attribute a failure correctly but fail to explicitly output the responsible step. Building on this finding, we propose \emph{grafting}, a simple manual intervention that exploits knowledge already encoded in the model to improve AFA performance without requiring ground-truth labels at inference time. Extensive evaluations across multiple LLMs and benchmark settings demonstrate that grafting consistently outperforms direct-prompting approaches, underscoring the potential of leveraging latent model knowledge for more reliable failure attribution. Our code is available at https://github.com/wendyzheng22/mi4afa.

View source

Similar papers

#artificial intelligence Preprint Oct 2026

Error-Propagation Modeling for Failure Attribution in LLM-Based Multi-Agent Systems

LLM-based multi-agent systems (MASs) are increasingly used to solve complex tasks through coordinated reasoning, tool use, and interaction with external resources. However, attributing failures in such systems remains challenging because the observed outcome often does not directly reveal the error responsible for the...

Jia-Qi Liao, Yuan-Zhao Zhai, Huan-Xi Liu et al. · 0 citations
Book Open access Nov 2026

TraJudger: Structure-Guided Failure Attribution for LLM-Powered Multi-Agent Systems

LLM-powered multi-agent systems are increasingly deployed in production, yet diagnosing their failures remains a major operational challenge. Unlike failures in cloud architectures, which can often be diagnosed using structured diagnostic information, multi-agent system failures are embedded in unstructured natural-lan...

Yong-Xin Zhao, Sheng-Lin Zhang, Wen-Wei Gu et al. · 0 citations

AgentBoundary: Counterfactual Evaluation of Safety in Tool-Using LLM Agents

This work introduces AgentBound, the first four-way counterfactual generation-and-evaluation framework for tool-using agent safety, and shows that effective agentic alignment requires action decisions to track permission-relevant execution evidence, rather than refusal strength alone.

Tian-Zhuo Yang, Zi-Rui Mi, Yan-Tao Huang et al. · 1 citation
Preprint Oct 2026

Testing the Planning Claims of LLM-Based Agents against Formal Semantics

LLM-based agents are increasingly credited with the competences of practical reasoning: planning, monitoring, and correcting their own actions. Their success is attributed to the prompting strategies and agent architectures built around the model, such as chain-of-thought, search-based prompting, self-refinement, refle...

Oriol Miro-Lopez-Feliu, Oscar Miró López-Feliu, Víctor Giménez-Ábalos et al. · 0 citations
#artificial intelligence Preprint Oct 2026

AlphaDiana: A Measurement System for Agents in Verifiable Reasoning Tasks

Agents pair reasoning models with harnesses that organize model calls, execute tools, load skill instructions, and retain history in memory. Understanding agent performance requires linking execution conditions to outcomes, behavior, and resource use. We introduce AlphaDiana, a measurement system with modular interface...

Zhan-Ke Zhou, Zong-Ze Li, Weikai Huang et al. · 3 citations
#artificial intelligence Preprint Oct 2026

HERA: Harness-Environment Co-Evolution for Reliable Agentic Abstention

Large language model (LLM) agents are increasingly capable of acting in complex tool-use environments, yet they often fail to recognize when tasks are infeasible and no valid solution exists. Recent work has formalized this reliability gap as the problem of agentic abstention, and existing approaches typically optimize...

Hang Luo, Bing-Bing Wen, Guang Yang et al. · 1 citation

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.