Search papers, labs, and topics across Lattice.
2
0
5
3
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.
Training on 500K automatically-curated ophthalmology instructions lets a vision-language model leapfrog general medical models in a specialized domain.
