Reproduce “Conversational Generalized Linear Bandits with Hybrid Feedback” 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

Yurui Zou, Qirun Zeng, Jinhang Zuo

HyConUCB combines item rewards with relative comparisons between key terms to learn user preferences. It uses both feedback types in one system. In a synthetic common-intercept setting, it has the lowest final mean cumulative regret at each tested intercept. It leads on Yelp, while Var-Random leads on the primary LLM-routing metric; conversation costs are not evaluated.

Summary by alphaXiv.

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