Reproduce “$τ$-Rec-Books: A Disclosure-Controlled Benchmark for Requirement Elicitation in Agentic Recommender Systems” 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
Bharath Sivaram Narasimhan, Karthik R. Narasimhan
τ-Rec-Books tests whether conversational recommenders can learn hard book requirements before recommending. It replays matched requests with requirements volunteered, revealed when asked, or partly withheld, and scores outcomes against catalog metadata. Across satisfiable tasks, average success falls from 0.975 to 0.756 to 0.283. These results concern simulated hard-filter tasks, not human conversational skill or book taste.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.