Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Affiliation:
1
0
0
Velocity penalties in RL reward functions are not just redundant for stabilization—they actively induce brittle policy gradient training as their coefficients scale.