Search papers, labs, and topics across Lattice.
School of Computing, University of Otago
2
0
3
Semantic-ID tokenizers, a promising approach for generative recommendation, can inflate Hit@K metrics by over 100% due to collisions where multiple items map to the same token sequence.
Achieve up to 17.6% recall and 16% NDCG gains in sequential recommendation by modeling transitions directly in the discrete semantic code space, effectively capturing fine-grained semantic dependencies often lost in aggregated item representations.
