Year: 2026
Type: preprint
Abstract: In many reinforcement learning (RL) domains, environments are connected by prerequisite relations, such as difficulty-increasing edits or parameter increments, which induce a directed acyclic curriculum graph (DAG). Although this structure is often exploited only implicitly, explicitly modeling it can improve training. We introduce PATH, a curriculum-learning framework that performs active learning over the curriculum graph. PATH first expands coverage by sampling diverse curriculum paths and then reallocates training toward regions that remain unmastered. Experiments across diverse environments show that PATH explicitly leverages the graph structure to achieve strong robustness and generalization.
Source: arXiv (Cornell University)
Authors: Liu, Zhenya, Chen, Yuxin
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: green
Awards: G1453181110, Track 4: Advanced CI Coordination Ecosystem: Monitoring and Measurement Services, ARTIFICIAL INTELLIGENCE (AI) – ASSISTED HYBRID RENEWABLE ENERGY, NUTRIENT, AND WATER RECOVERY FROM MUNICIPAL WASTEWATER, NSF-Simons AI Institute for the Sky (SkAI Institute), G3562386289, G3927959030, G3993056153, Track 1: ACCESS Resource Allocations Marketplace and Platform Services (RAMPS), G4010157413, ACO: An Open CI Ecosystem to Advance Scientific Discovery (OpenCI), Track 3: COre National Ecosystem for CyberinfrasTructure (CONECT), G5098925467, G5367127336, G5464919783, G6247045938, Collaborative Research: RI: Small: Active Curriculum and Environment Design for Reinforcement Learning, G754761654, Track 2: Customized Multi-tier Assistance, Training, and Computational Help (MATCH) for End User ACCESS to CI, Collaborative Research: RI: Small: Foundations of Few-Round Active Learning
OpenAlex ID: W7204520017
Cites:
Cited by:
Related to: