TRAJDEBUG: Tracing Error Lifecycle to Identify Critical Failures in Long-Horizon Agent Trajectories

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Yunjia Qi, Zehua Yin, Xintong Shi, Hao Peng, Songyuanyi Lu, Yixian Liu, Richeng Xuan, Yuhong Liu, Zhichao Hu, Xiaozhi Wang, Lei Hou, Bin Xu, Juanzi Li
• Venue: arXiv: cs.AI
• Published: 2026-08-06
• Collected: 2026-08-07
• URL: http://arxiv.org/abs/2608.06346v1
• PDF: https://arxiv.org/pdf/2608.06346v1
• DOI: UNKNOWN

Plain-Language Brief

The paper presents TrajDebug, a framework designed to trace error lifecycles in long-horizon AI agents. The goal is to identify critical failures by compressing history and using evidence-based error identification. The framework was tested on various AI benchmarks and outperformed existing baselines. The authors plan to release the codes and data for further research.

What the Document Says

• The paper introduces TrajDebug, an error-lifecycle tracing framework for long-trajectory error discovery in LLM-based agentic systems.
• TrajDebug addresses difficulties in identifying individual errors in long trajectories and attributing the final failure to specific errors.
• It uses multi-granularity history compression and evidence-based error identification for long-trajectory error discovery.
• The framework supports critical attribution by tracing each error's resolution status and terminal impact.
• A benchmark, TrajErrBench, was constructed with 486 manually annotated failed trajectories from Tau2Bench and SWE-Bench Pro for realistic tool-use and coding scenarios.
• Experiments showed that TrajDebug outperformed existing baselines across diverse agent benchmarks.
• Application studies demonstrated that TrajDebug's diagnoses provide actionable feedback for improving downstream agent success.
• The authors plan to release the codes and data for further research.

Key Concepts Mentioned

• TrajDebug
• error lifecycle
• LLM-based agentic systems
• critical error detection
• multi-granularity history compression
• evidence-based error identification
• Tau2Bench
• SWE-Bench Pro
• TrajErrBench

Claims Made by the Document

• TrajDebug addresses two main challenges in critical error detection.
• TrajDebug outperformed existing baselines across diverse agent benchmarks.
• Application studies demonstrate that TrajDebug's diagnoses provide actionable feedback.

Evidence Used

• Results from experiments on various AI benchmarks

Methods / Approach

• Multi-granularity history compression
• Evidence-based error identification

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

LLM-based agentic systems have shown remarkable capabilities in complex domains, while suffering from cascading errors and difficulty in debugging. Critical error detection aims to locate the earliest error step in a failed trajectory that is responsible for the final failure. However, progress faces two main challenges. First, long trajectories make it difficult to identify individual errors, since the evidence for judging a step may be scattered across distant instructions, observations, and prior context. Second, failed trajectories often contain multiple local errors with different downstream effects, only some of which remain responsible for the final failure. In this work, we propose TrajDebug, an error-lifecycle tracing framework that addresses long-trajectory error discovery with multi-granularity history compression and evidence-based error identification, and supports critical attribution by tracing each error's resolution status and terminal impact. We further construct TrajErrBench, a benchmark of 486 manually annotated failed trajectories from Tau2Bench and SWE-Bench Pro, covering realistic tool-use and coding scenarios. Experiments across diverse agent benchmarks show that TrajDebug achieves the best overall performance over existing baselines, and application studies further demonstrate that its diagnoses provide actionable feedback for improving downstream agent success. We will release the codes and data to facilitate further research.