Towards General Language-Conditioned Latent Safety Filters

2026-07-31

AI Summary

This work proposes a language-conditioned safety control framework that directly integrates natural language constraints into Hamilton-Jacobi-based safety-critical control, addressing the limited generalizability of existing robotic safety filters tailored to specific tasks or predefined constraints. By embedding linguistic specifications into the latent space of a vision-language-action framework, the method introduces a novel language-conditioned Actor-Critic architecture that dynamically enforces safety constraints during execution. The approach enables partial generalization to unseen yet semantically related instructions, significantly reducing constraint violations across diverse manipulation tasks—including pick-and-place, tabletop wiping, and block stacking—while demonstrating strong language-conditioned transferability and broad adaptability to varied safety requirements expressed in natural language.

Abstract

Robot policies are becoming increasingly general, with vision-language-action (VLA) models enabling a single policy to execute diverse tasks specified in natural language. Safe deployment, however, requires adapting not only to new tasks but also to varying safety requirements across users, environments, and applications. Existing safety filters remain largely constraint-specific and thus must be redesigned or relearned when safety requirements change. In this paper, we investigate language-conditioned safety filtering, in which a Hamilton-Jacobi safety actor and critic are conditioned on language-specified constraints. We evaluate this formulation across pick-and-place, table-wiping, and block-stacking tasks in the vision-based setting, examining its ability to enforce language-specified constraints and transfer to unseen constraint instances within the evaluated constraint families. Our experiments provide evidence that language-conditioned safety filters reduce constraint violations and exhibit partial transfer to unseen constraint instances.

Problem

• safety filtering
• language-conditioned constraints
• robot safety
• generalization
• constraint adaptation

Innovation

• language-conditioned safety
• Hamilton-Jacobi reachability
• vision-language-action models
• safety filtering
• constraint generalization