Improving task success by fixing planner and actor state disagreements
Consistent Plan-Act for Long-Horizon Agentic Tasks
Artificial Intelligence
Summary
Planning and acting in complicated tasks can fail when the planner and actor don’t agree on the task’s state. The authors found that these disagreements cause coordination problems and reduce success. They showed that giving both planner and actor shared, relevant state facts helps reduce confusion. They developed a system called Consistent Plan-Act that detects and resolves contradictions between planner and actor states, which improves task performance in test environments.
What this means in practice
- •For robotics developers: Enhance robot task execution by synchronizing planning and acting modules to reduce misunderstandings about the environment state.
- •For game ai engineers: Increase success rates of AI agents in simulated environments by feeding back inconsistencies between planning and action stages for better coordination.
Authors
Heng-Zhuang Li, Yi-Kai Zhang, Yu Wang, Yueqing Sun, Jiayuan Zhang, Qi Gu, Han-Jia Ye
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.