Search papers, labs, and topics across Lattice.
Affiliation:
1
0
2
0
A single feedforward network learns to predict safety values and amortize minimal-intervention filtering, enabling safe RL in high-dimensional systems without analytical safety functions or competing policy objectives.
