Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Affiliation:
2
0
3
A scalar adjoint method slashes the computational cost of fine-tuning flow policies in RL, achieving state-of-the-art results on complex manipulation tasks.
TRQAM stabilizes off-policy reinforcement learning by precisely controlling deviations from pretrained policies, leading to a 68% success rate—22% higher than the best prior method.