Differentiable Learning of Lifted Action Schemas for Classical Planning

2026-05-13 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Expert, quick

Summary

A novel neural network architecture has been developed for learning lifted action schemas in classical planning domains. This architecture addresses the challenge of inferring action schemas and their arguments from observed state changes, specifically when action arguments are unobserved but states are fully observed. The work aims to provide a robust differentiable component for integration into larger neuro-symbolic models, representing a crucial step towards learning planning domains from sequences of images and action labels. The architecture's effectiveness is evaluated across various planning domains, demonstrating its ability to recover ground-truth action schema structures. Experiments also assess its robustness to observation noise and its performance in a slot-based dynamics model variation.

Key takeaway

For research scientists developing neuro-symbolic AI, this architecture offers a robust method for learning planning domain dynamics from incomplete observations. You should consider integrating this differentiable component to improve the generalization and interpretability of your models, especially when dealing with scenarios where action arguments are not directly observable but state transitions are.

Key insights

A neural network learns lifted action schemas and arguments from observed state changes with unobserved actions.

Principles

• Lifted action schemas enable structural generalization.
• Differentiable components integrate into neuro-symbolic models.

Method

The approach uses a neural network to simultaneously learn action schemas and identify action arguments from fully observed state changes, even when action arguments themselves are unobserved.

In practice

• Integrate into larger neuro-symbolic models.
• Apply to planning domains with unobserved action arguments.

Topics

• Differentiable Learning
• Lifted Action Schemas
• Classical Planning
• Neural Network Architecture
• Action Argument Identification

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, AI Engineer

Related on AIssential

• Differentiable Learning of Lifted Action Schemas for Classical Planning — A novel neural network learns lifted action schemas from state traces with unobserved action arguments.
• Scaling Neuro-symbolic Problem Solving: Solver-Free Learning of Constraints and Objectives — A neuro-symbolic architecture with a probabilistic loss efficiently learns to solve NP-hard problems from natural inputs.
• Improving Vision-Language-Action Model Fine-Tuning with Structured Stage and Keyframe Supervision — Structured supervision significantly enhances Vision-Language-Action model fine-tuning for robotic manipulation.
• Beyond KV-Cache: Test-Time Training & Invariant Latent Topologies for ICL — New methods enable LLMs to learn during inference and generalize across diverse domains by modifying internal weight structures or identifying invariant neural subspaces.
• Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability — Reasoning SFT can generalize cross-domain, but it is conditional on optimization, data quality, and model capability.
• Lifting Embodied World Models for Planning and Control — Lifting world models with low-dimensional waypoints significantly improves embodied agent planning efficiency and accuracy.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.