Search papers, labs, and topics across Lattice.
3
0
5
0
DQC reduces the mean-collapse issue in multimodal regression, achieving a remarkable test minMSE of 0.19鈥攆ar superior to random and mean collapse strategies.
Orthogonalized optimization with the Muon optimizer can significantly boost adversarial robustness, outperforming AdamW on both CNNs and ViTs, highlighting the importance of optimizer geometry in security-critical training.
SkillOpt transforms agent skill development into a reproducible optimization process, achieving state-of-the-art results by treating skills as trainable parameters.
