Saved in:
Bibliographic Details
Main Authors: Chen, Mengzhuo, Wang, Junjie, Mu, Fangwen, Wang, Yawen, Liu, Zhe, Feng, Huanxiang, Wang, Qing
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.22708
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914505187917824
author Chen, Mengzhuo
Wang, Junjie
Mu, Fangwen
Wang, Yawen
Liu, Zhe
Feng, Huanxiang
Wang, Qing
author_facet Chen, Mengzhuo
Wang, Junjie
Mu, Fangwen
Wang, Yawen
Liu, Zhe
Feng, Huanxiang
Wang, Qing
contents Failure attribution, i.e., identifying the responsible agent and decisive step of a failure, is particularly challenging in LLM-based multi-agent systems (MAS) due to their natural-language reasoning, nondeterministic outputs, and intricate interaction dynamics. A reliable benchmark is therefore essential to guide and evaluate attribution techniques. Yet existing benchmarks rely on partially observable traces that capture only agent outputs, omitting the inputs and context that developers actually use when debugging. We argue that failure attribution should be studied under full execution observability, aligning with real-world developer-facing scenarios where complete traces, rather than only outputs, are accessible for diagnosis. To this end, we introduce TraceElephant, a benchmark designed for failure attribution with full execution traces and reproducible environments. We then systematically evaluate failure attribution techniques across various configurations. Specifically, full traces improve attribution accuracy by up to 76\% over a partial-observation counterpart, confirming that missing inputs obscure many failure causes. TraceElephant provides a foundation for follow-up failure attribution research, promoting evaluation practices that reflect real-world debugging and supporting the development of more transparent MASs.
format Preprint
id arxiv_https___arxiv_org_abs_2604_22708
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Seeing the Whole Elephant: A Benchmark for Failure Attribution in LLM-based Multi-Agent Systems
Chen, Mengzhuo
Wang, Junjie
Mu, Fangwen
Wang, Yawen
Liu, Zhe
Feng, Huanxiang
Wang, Qing
Multiagent Systems
Failure attribution, i.e., identifying the responsible agent and decisive step of a failure, is particularly challenging in LLM-based multi-agent systems (MAS) due to their natural-language reasoning, nondeterministic outputs, and intricate interaction dynamics. A reliable benchmark is therefore essential to guide and evaluate attribution techniques. Yet existing benchmarks rely on partially observable traces that capture only agent outputs, omitting the inputs and context that developers actually use when debugging. We argue that failure attribution should be studied under full execution observability, aligning with real-world developer-facing scenarios where complete traces, rather than only outputs, are accessible for diagnosis. To this end, we introduce TraceElephant, a benchmark designed for failure attribution with full execution traces and reproducible environments. We then systematically evaluate failure attribution techniques across various configurations. Specifically, full traces improve attribution accuracy by up to 76\% over a partial-observation counterpart, confirming that missing inputs obscure many failure causes. TraceElephant provides a foundation for follow-up failure attribution research, promoting evaluation practices that reflect real-world debugging and supporting the development of more transparent MASs.
title Seeing the Whole Elephant: A Benchmark for Failure Attribution in LLM-based Multi-Agent Systems
topic Multiagent Systems
url https://arxiv.org/abs/2604.22708