Search papers, labs, and topics across Lattice.
1
0
3
MLLMs can be significantly boosted by curriculum learning that focuses on reward design rather than data selection, dynamically weighting generalized rubrics based on the model's evolving competence.
