Search papers, labs, and topics across Lattice.
4
0
6
By strategically amplifying updates along flat directions in the loss landscape, LITE unlocks faster LLM pre-training with existing matrix-based optimizers like Muon and SOAP.
Perturbations in forward and backward passes of SGD can cascade geometrically through deep networks, but this paper identifies conditions under which asymptotic convergence order is preserved.
LLMs can generate unbiased pseudo-labels for unexposed items in pre-ranking, boosting click-through rate by 3.07% in production while improving diversity.
Taobao's recommender system just got a 1.65% CTR boost by compressing ultra-long user behavior sequences with a hierarchical codebook and sparse attention, proving that personalized interest centers can be learned efficiently.
