Reproduce “OptST: Optimizing Semantic Tokenization for LLM-Enhanced Recommendation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hanchong Chen, Jie Liu, Fuyuan Lyu, Dugang Liu, Lingjie Li, Xiuqiang He, Xing Tang
OptST adapts a pretrained semantic tokenizer for recommendation by learning low-rank transformations of its codebooks while keeping the encoder and base codebooks frozen. It also selects a user–item-specific prefix of semantic tokens. Across four public offline rating datasets, OptST achieves the highest AUC among 13 reported baselines, with significant gains over the strongest baseline; LogLoss results are less uniform.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.