Get Started
Home
Topics
Search
Library
Research questionHow can production recommender systems continually optimize retrieval, ranking, and serving as users and content change?Production recommenders can become misaligned with changing content and user behavior, leaving retrieval, ranking, and serving decisions outdated. Revising these decisions manually through online experiments is slow and difficult to sustain.
AI
AI Agents
Inference Optimization
Information Retrieval
Machine Learning
Research Paper
Technology
Latest papersRecent research connected to this question, newest first.CORAL: An LLM-Native Harness for Production Recommender SystemsThe evidence concerns CORAL, an LLM-based closed-loop harness that observes operating signals, uses decision history and tools, and reconfigures live recommenders within a fixed operating budget without parameter updates. Results come from A/B experiments on two large-scale social platforms, showing engagement or serving-cost improvements without additional serving cost or degraded engagement in the reported settings.research paper · Sep 2, 2026
Related questions
How can industrial recommender teams reliably coordinate long-running experiments from research through launch review?How can recommenders adapt to changing preferences without trusting unreliable inputs?How can production RAG teams maintain reliable comparisons as new retrieval candidates arrive without rejudging overlapping documents?How can machine-learning performance-modeling frameworks stay maintainable as models and systems evolve?