Reproduce “TransX Technical Report: Scaling Sequence Transduction for 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
Da Xu, Liyan Fang, Divya Venugopalan, Sunny Hsu, Xukai Wang, Rishav Roy Chowdhury, Cindy Liang, Nishant Satya Lakshmikanth
TransX separates history encoding from candidate-specific scoring in recommendation. It uses a causal encoder and bounded memory that preserves recent behavior in detail and summarizes older behavior, which candidates query to make predictions. In production, TransX-Base improved CTR by 6.02% and conversion by 4.45% versus a sequential baseline; normalized disaggregated total serving cost was 0.22 of colocated cost in this setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.