Reproduce “Beyond Preference Matching: Towards Need-Aware Personalized Large Language Models” 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

Pradeep Bolleddu, Kamal Chilukuri, Alisa Banerjee

The authors test whether checking for a mismatch between a user’s request and inferred need can improve responses. In a prompted strategy, GPT-4o analyzes prompts and answers to the inferred need; comparisons use simulated user preferences. Across 1,297 divergent cases, preference-matched responses score higher on preference reward, while need-aware responses score higher on need alignment, not establishing users’ needs.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.