Reproduce “Optimizing Sparse Outcomes Through Dense Behavioral Signals via Value-Guided Preference Distillation” 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
Ziyi Zhu, Daniel R. Cahn, Thomas D. Hull, Caitlin A. Stamatis, Olivier Tieleman, Guilherme B. Freire, Jinghong Chen
The paper presents a chatbot method that predicts multiple user behaviors across message and day horizons, combines these signals to select responses, and distills the selection policy into the chatbot. In a three-week A/B test with 6,700 users per arm, the deployed policy increased D7 retention by 2.60 percentage points. This was one mental-health product, and retention is not symptom improvement.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.