Research map / Reinforcement learning / Offline RL

Safe Offline Reinforcement Learning using Behavior Regularisation and Latent Feasibility-Guidance

NeurIPS 2026 · Reinforcement learning › Offline RL · approach: Pessimism and uncertainty in offline RL

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes an offline safe RL method combining behavior regularisation with latent feasibility guidance to enforce safety constraints in a latent-variable policy space rather than via soft constraints.

Problem

Offline safe reinforcement learning must maximize reward under safety constraints using only static datasets; existing latent-space methods rely on soft constraints that allow expected constraint violations.

Approach

Uses behavior regularisation together with latent feasibility guidance in a latent variable model of the behavior policy, replacing soft constraints with guidance that targets feasible actions (details beyond the abstract not available).

Closest papers by meaning

• Latent Safety-Constrained Policy Approach for Safe Offline Reinforcement Learning (ICLR 2025 · similarity 0.94)
• Safe Offline Reinforcement Learning with Feasibility-Guided Diffusion Model (ICLR 2024 · similarity 0.92)
• Offline Safe Reinforcement Learning Using Trajectory Classification (AAAI 2025 · similarity 0.92)
• Constraint-Adaptive Policy Switching for Offline Safe Reinforcement Learning (AAAI 2025 · similarity 0.91)
• Adaptive Advantage-Guided Policy Regularization for Offline Reinforcement Learning (ICML 2024 · similarity 0.91)
• A2PO: Towards Effective Offline Reinforcement Learning from an Advantage-aware Perspective (NeurIPS 2024 · similarity 0.91)