Chapter 3 · 2026
AgentRx: Diagnosing AI Agent Failures from Execution Trajectories
Shraddha Barke, Arnav Goyal, Alind Khare
Abstract
AI agents often fail in ways that are difficult to localize due to probabilistic, long-horizon, multi-agent executions and noisy tool outputs. This paper addresses this by manually annotating failed agent runs and releasing a novel benchmark of 115 failed trajectories. It also presents AGENTRX, an automated diagnostic framework that pinpoints critical failure steps and categories, improving step localization and failure attribution over existing baselines.
Topics
AI agent failuresdiagnosisexecution trajectoriesmulti-agent systemserror localizationfailure attribution
Relevance Scores
Long-Horizon Score65
Enterprise Score60
Completeness75
Paper Info
Year2026
Venue
Type
ChapterCh. 3
Authors3
Zone III Analysis
Related Papers
Tree of Thoughts: Deliberate Problem Solving with Large…
2023 · Ch.1
HuggingGPT: Solving AI Tasks with ChatGPT and its Frien…
2023 · Ch.4
Generative Agents: Interactive Simulacra of Human Behav…
2023 · Ch.2
MemGPT: Towards LLMs as Operating Systems
2023 · Ch.2
View all Chapter 3 papers →