Consistent Plan-Act for Long-Horizon Agentic Tasks

2026-09-29

AI Summary

This study addresses the coordination failures between planning and execution agents in long-horizon tasks caused by inconsistent state cognition. To mitigate this issue, we propose ConPAct, a framework that localizes state conflicts through structured state assertions and programmatic contradiction detection. Furthermore, it integrates consistency-interactive fine-tuning to enable inference-time correction and collaborative training. Experimental results demonstrate that ConPAct improves task success rates on MiniGrid from 38.6% to 54.4% while significantly enhancing cross-environment generalization capabilities. By effectively aligning state representations across heterogeneous agents, this work establishes a novel paradigm for efficient multi-agent collaboration in complex sequential decision-making environments.

Abstract

Long-horizon agentic tasks demand strong reasoning and efficient execution across successive interactions with dynamic environments. A common approach decouples high-level planning from low-level execution through separate planner and actor roles. To investigate coordination failures in these tasks, we prompt both agents for structured state assertions and compare their reports programmatically to detect explicit contradictions. Our analyses reveal systematic disagreement about the same task-relevant state facts, a phenomenon we term planner-actor state mismatch. We further find that providing agents with task-relevant state information reduces mismatch and improves coordination and task performance. Based on the systematic analysis of the state mismatch, we propose Consistent Plan-Act (ConPAct), which feeds detected contradictions back to both agents to form consistent state interpretations and fine-tunes them on curated consistent interactions for better coordination. ConPAct improves performance across various environments and model configurations, e.g., increasing MiniGrid success rate from 38.6% to 54.4% with GPT-5.6-sol/terra as planner and actor respectively, demonstrating that state consistency can guide both inference-time correction and coordination training.

Problem

• long-horizon agentic tasks
• planner-actor coordination
• state mismatch
• multi-agent systems

Innovation

• Planner-Actor State Mismatch
• Consistent Plan-Act (ConPAct)
• Long-Horizon Agentic Tasks
• State Assertions
• Coordination Fine-tuning