AgentRewind: Recoverable Execution for Long-Horizon LLM Agents

2026-08-14

AI Summary

This study addresses the challenge of irreversible early errors in long-horizon LLM agents by proposing a runtime recovery framework. We introduce a novel context-and-environment-aligned checkpointing mechanism that enables state rollback and historical information reuse, effectively preventing error propagation. Furthermore, we construct MettleBench, an evaluation benchmark that validates our approach through comprehensive multi-model comparisons. Experimental results demonstrate that the proposed framework significantly improves task success rates and checklist progress, outperforming existing baselines. These findings establish a new paradigm for ensuring robust execution in long-horizon autonomous agents, offering a scalable solution to mitigate cascading failures in complex agentic workflows.

Abstract

Many real-world tasks require LLM agents to interact with their environments over long execution horizons. Errors that occur early in execution may propagate through both the agent context and environment state, and their effects may be difficult to reverse through subsequent actions. Existing methods mainly seek to reduce such errors through plan refinement and safety checks but provide little support after errors occur. To enable recovery during long-horizon execution, we present AgentRewind, a runtime recovery framework that records aligned checkpoints of the agent context and controlled environment, allowing agents to return to an earlier state and resume execution with information from previous attempts. We also construct MettleBench, a benchmark for evaluating task completion and partial progress on long-horizon engineering assignments containing a series of related requirements. Experiments across tasks, multiple models, execution strategies, and agent harnesses show that AgentRewind improves task success rate and average checklist progress over the compared baselines.

Problem

• Long-Horizon LLM Agents
• Error Recovery
• Error Propagation
• Runtime Recovery

Innovation

• AgentRewind
• Recoverable Execution
• Aligned Checkpoints
• Long-Horizon LLM Agents
• MettleBench