Learning General Policies From Examples

2025-09-02

AI Summary

Existing combinatorial planning approaches generate verifiable general policies but are constrained by SAT/ASP frameworks, limiting scalability to only hundreds of states and features. This paper proposes a novel symbolic learning paradigm for sample-based policy generalization, abandoning traditional constraint solving in favor of hitting-set-driven efficient symbolic induction, augmented with structured termination and acyclicity guarantees. Our method achieves, for the first time, scalable policy learning over state spaces of up to one million states and feature spaces of up to one hundred thousand features. It significantly outperforms existing symbolic planners across multiple benchmarks while ensuring strong interpretability, formal correctness (via verification), and practical scalability. By bridging symbolic reasoning with scalable learning, our approach establishes a new pathway for learning general-purpose planning policies in large-scale domains.

Technology Category

• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Combinatorial methods for learning general policies that solve large collections of planning problems have been recently developed. One of their strengths, in relation to deep learning approaches, is that the resulting policies can be understood and shown to be correct. A weakness is that the methods do not scale up and learn only from small training instances and feature pools that contain a few hundreds of states and features at most. In this work, we propose a new symbolic method for learning policies based on the generalization of sampled plans that ensures structural termination and hence acyclicity. The proposed learning approach is not based on SAT/ASP, as previous symbolic methods, but on a hitting set algorithm that can effectively handle problems with millions of states, and pools with hundreds of thousands of features. The formal properties of the approach are analyzed, and its scalability is tested on a number of benchmarks.

Problem

• Learning scalable general policies from examples
• Overcoming limitations of small state and feature pools
• Ensuring structural termination and acyclicity in policies

Innovation

• Symbolic method using plan generalization
• Hitting set algorithm replacing SAT/ASP
• Scales to millions of states and features