Executable Prefix Values for Locating Agent Failures

cs.AI, cs.CL

Submitted: 2026-10-13

Updated: 2026-10-13

Comments: 46 pages, 12 figures, 60 tables

Code: https://github.com/SWE-agent/mini-swe-agent

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• Thinking Before Thinking: Scaling Agentic Inference Through Meta-Reasoning
• SWE-bench: Can Language Models Resolve Real-World GitHub Issues?
• DoVer: Intervention-Driven Auto Debugging for LLM Multi-Agent Systems
• Locating Hidden Failures Makes Long-Horizon Agents More Reliable
• Fail-Fast, Restart-Smart: Early Failure Prediction and Restart for SWE Agentic Tasks
• SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering
• Which Agent Causes Task Failures and When? On Automated Failure Attribution of LLM Multi-Agent Systems
• LongRCA Bench: Root-Cause Localization in Long-Horizon Agent Trajectories
• Where LLM Agents Fail and How They can Learn From Failures
• Rollout Pass-Rate Control: Steering Binary-Reward RL Toward Its Most Informative Regime

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems