Reproduce “HASTE: Hierarchical and Attribute-aware Supervision for Generative 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
Tripti Shukla, Shwetha Somasundaram, Akshat Gupta, Rohit Garg, Paul Duff, Ian Anderson
HASTE trains generative recommenders to predict semantic item IDs while weighting coarse codes more heavily and adding catalog-attribute prediction from hidden states. Confidence-gated losses and sampled prefixes account for earlier-code uncertainty. On Fashion-140k’s Tail split, adding color supervision raises Recall@10 from 0.0215 to 0.0267; hierarchical loss also reaches the CE validation-loss target in 1.99× fewer steps, not elapsed time.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.