Search papers, labs, and topics across Lattice.
1
0
3
Forget expensive human annotations: RubiCap uses LLM-generated rubrics to train image captioning models via RL, achieving superhuman performance and even improving VLM pretraining.
