Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning

arXiv:2609.34575 · cs.AI · Submitted 2026-09-28 · Read on arXiv

cs.AI

Submitted: 2026-09-28

Updated: 2026-09-28

Code: https://github.com/Henry0132/DSP

Project page: https://henry0132.github.io/DSP

Terminology

• Reinforcement Learning — 1× in this paper · explained in 4 episodes

Sources

• Option-aware Temporally Abstracted Value for Offline Goal-Conditioned Reinforcement Learning
• Is Conditional Generative Modeling all you need for Decision-Making?
• Simple Hierarchical Planning with Diffusion
• Score Regularized Policy Optimization through Diffusion Behavior
• Chain-of-Goals Hierarchical Policy for Long-Horizon Offline Goal-Conditioned RL
• Diffusion Guidance Is a Controllable Policy Improvement Operator
• Learning to Reach Goals via Iterated Supervised Learning
• Physics-informed Value Learner for Offline Goal-Conditioned Reinforcement Learning
• IDQL: Implicit Q-Learning as an Actor-Critic Method with Diffusion Policies
• Classifier-Free Diffusion Guidance
• Planning with Diffusion for Flexible Behavior Synthesis
• Prior-Guided Diffusion Planning for Offline Reinforcement Learning
• Offline Goal-Conditioned Reinforcement Learning with Projective Quasimetric Planning
• Offline Reinforcement Learning with Implicit Q-Learning
• Flow Matching for Generative Modeling
• Goal-Conditioned Reinforcement Learning: Problems and Solutions
• SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
• Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow
• Stochastic Interpolants via Conditional Dependent Coupling
• Drift Flow Matching

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems