Search papers, labs, and topics across Lattice.
1
0
3
Forget auxiliary encoders and handcrafted losses: LVRPO uses reinforcement learning to directly align language and vision, boosting performance across a range of multimodal tasks.
