Johnson Lui

Tokenizing Numerical and Embedding Features for LLM RecSys

Zhe Xu, Ankit Peshin, Chiyu Zhang, Feng Qi, Johnson Lui, Anil Ramakrishna, et al.

Large language models (LLMs) are increasingly used as backbone architectures for recommender systems because of their strong sequence modeling and representation learning capabilities. However, most LLM-based recommenders operate primarily on discrete textual tokens, whereas prac…Read more