Reproduce “Online LLM Personalization from Pairwise Preferences Across Users” 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
Romain Puech, Georgy Salakhutdinov, Dimitris Bertsimas
CODuel personalizes tutoring for each new student by choosing between two answer styles across a short session. It fits preference patterns from completed sessions to initialize a belief for each newcomer, then updates that belief from their choices. On held-out simulated profiles, CODuel scores 48 versus 30 for its no-transfer ablation, but these results do not establish improved satisfaction or learning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.