Reproduce “Raw Histories or Compact Profiles? A Cost-Aware Empirical Study of User History Representations for Personalized Review Generation with LLMs” 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
Naofumi Osawa
The authors test compact user profiles for personalized review drafting, comparing them with prompts containing users’ past reviews. Profiles summarize writing style, preferences, title style, and typical structure before generation. In Amazon Fashion, GPT-5.4 with up to 30 reviews scores higher on all four automatic metrics with raw histories; profile reuse lowers recurring estimated cost only when history is reused.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.