ROAIFeb 27

SafeGen-LLM: Enhancing Safety Generalization in Task Planning for Robotic Systems

Jialiang Fan, Weizhe Xu, Mengyu Liu, Oleg Sokolsky, Insup Lee, Fangxin Kong
arXiv:2602.24235v1
Originality Incremental advance
AI Analysis

This addresses safety generalization in task planning for robotic systems, representing an incremental improvement over existing methods.

The paper tackles the challenge of safety-critical task planning in robotic systems by proposing SafeGen-LLM, which enhances safety satisfaction and generalizes to novel safety properties, achieving strong safety generalization and outperforming proprietary baselines across multi-domain planning tasks.

Safety-critical task planning in robotic systems remains challenging: classical planners suffer from poor scalability, Reinforcement Learning (RL)-based methods generalize poorly, and base Large Language Models (LLMs) cannot guarantee safety. To address this gap, we propose safety-generalizable large language models, named SafeGen-LLM. SafeGen-LLM can not only enhance the safety satisfaction of task plans but also generalize well to novel safety properties in various domains. We first construct a multi-domain Planning Domain Definition Language 3 (PDDL3) benchmark with explicit safety constraints. Then, we introduce a two-stage post-training framework: Supervised Fine-Tuning (SFT) on a constraint-compliant planning dataset to learn planning syntax and semantics, and Group Relative Policy Optimization (GRPO) guided by fine-grained reward machines derived from formal verification to enforce safety alignment and by curriculum learning to better handle complex tasks. Extensive experiments show that SafeGen-LLM achieves strong safety generalization and outperforms frontier proprietary baselines across multi-domain planning tasks and multiple input formats (e.g., PDDLs and natural language).

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes