Search papers, labs, and topics across Lattice.
4
0
9
10
Because mode-seeking reverse KL aggressively amplifies incorrect teacher signals, gating dense distillation on verifier-scored teacher probes systematically outperforms uniform distillation while reclaiming massive amounts of idle teacher compute.
A compact 1.7B parameter policy can outperform larger models in recommendation tasks by leveraging simulated user feedback for training.
RefVideo-6M revolutionizes video editing datasets by providing 5 million high-quality editing samples that prioritize visual references over flawed automatic edits.
Skill Optimizers trained through execution feedback can outperform traditional models by over 9 points, revealing a critical gap in agent learning methodologies.
