Search papers, labs, and topics across Lattice.
Huawei Technologies
3
0
6
High aliasing rates in Semantic-ID tokenizers can be detected before training, revealing that deterministic prefix controls significantly outperform learned mappings in alignment.
Identifying and resolving same-capability ambiguity can drastically improve agent safety, with SkillResolve achieving a remarkable 0% harmful sibling exposure.
Generative recommenders get a major upgrade: HPGR leverages hierarchical pre-training and sparse attention to dramatically improve performance and efficiency by explicitly modeling the structure of user behavior.
