Reproduce “MOSS: A Production Framework for Monitoring and Diagnosing Semantic Failures in Search and Recommendation 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
Alina Nesen, Prathyusha Kanmanth Reddy, David Relyea, Rohit Avinash Mujumdar, Samaneh Karimi, Bo Shen, Jordan Hashemi
MOSS uses an LLM to score sampled search, recommendation, and offer results, then diagnoses low-scoring cases for human review. In production, it ran for over six months and surfaced two confirmed serving defects. In an offline paired comparison, an enriched bundle placement reduced the pair-level alert rate from 41.6% to 9.8%, measuring a proposed fix rather than online impact.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.