SAS achieves test-time safety adaptation in offline RL by using Lyapunov-guided imagined trajectories recycled as in-context prompts in a transformer agent.
[Yes] Justification: The pa- per cites D4RL, RL-Unplugged, DSRL, and Safety Gymnasium
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Lyapunov-Guided Self-Alignment: Test-Time Adaptation for Offline Safe Reinforcement Learning
SAS achieves test-time safety adaptation in offline RL by using Lyapunov-guided imagined trajectories recycled as in-context prompts in a transformer agent.