Active Epistemic Control for Query-Efficient Verified Planning

2026-02-03

AI Summary

This work addresses the challenge of planning under partial observability, where task-critical preconditions are unknown and information acquisition is costly, rendering models prone to infeasible plans due to prediction errors. To this end, the paper proposes an Active Epistemic Control (AEC) framework that introduces, for the first time, a cognitive-categorical planning layer. This layer enforces a strict separation between belief and feasibility certification by maintaining distinct grounded fact and belief stores: beliefs are used solely for efficiency optimization, while feasibility is guaranteed through categorical checks and SQ-BCP pullback compatibility verification. Integrating model-based belief tracking with uncertainty-driven active querying, AEC efficiently prunes candidate plans. Evaluated on ALFWorld and ScienceWorld, the approach achieves success rates comparable to strong LLM agents while requiring significantly fewer replanning cycles.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Multiagent Systems: Multiagent Planning
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Planning in interactive environments is challenging under partial observability: task-critical preconditions (e.g., object locations or container states) may be unknown at decision time, yet grounding them through interaction is costly. Learned world models can cheaply predict missing facts, but prediction errors can silently induce infeasible commitments. We present \textbf{Active Epistemic Control (AEC)}, an epistemic-categorical planning layer that integrates model-based belief management with categorical feasibility checks. AEC maintains a strict separation between a \emph{grounded fact store} used for commitment and a \emph{belief store} used only for pruning candidate plans. At each step, it either queries the environment to ground an unresolved predicate when uncertainty is high or predictions are ambiguous, or simulates the predicate to filter hypotheses when confidence is sufficient. Final commitment is gated by grounded precondition coverage and an SQ-BCP pullback-style compatibility check, so simulated beliefs affect efficiency but cannot directly certify feasibility. Experiments on ALFWorld and ScienceWorld show that AEC achieves competitive success with fewer replanning rounds than strong LLM-agent baselines.

Problem

• partial observability
• verified planning
• query efficiency
• epistemic uncertainty
• feasibility certification

Innovation

• Active Epistemic Control
• verified planning
• partial observability
• belief management
• query-efficient