Publication
Safe Continual Reinforcement Learning under Nonstationarity via Adaptive Safety Constraints
Summary
Presents LILAC+, a predictive safety layer combining context-based, adaptation-speed, and soft-to-hard safety constraints for policy-level adaptation and runtime enforcement.
safe continual reinforcement learningadaptive safety constraintsnonstationarityLILAC+runtime safety
Citation information
Timofey Tomashevskiy. “Safe Continual Reinforcement Learning under Nonstationarity via Adaptive Safety Constraints.” 2026. arXiv:2605.18842.
The author-name form above follows the corresponding publication/preprint record. Tim Tomashevskiy also publishes under Timofey Tomashevskiy and Timothy Tomashevskiy.
Related research
This work forms part of a broader research program on adaptive AI safety, safe continual reinforcement learning, nonstationarity, and runtime assurance.