Search papers, labs, and topics across Lattice.
1
0
3
4
Mags-RL lets multimodal LLMs see the forest *and* the trees, using reinforcement learning to guide a super-resolution agent that selectively enhances image regions for improved reasoning without extra annotations.
