Tag
Jim Nielsen writes a response to Marques Brownlee's video essay 'Dear YouTube', arguing that YouTube's new A/B testing feature for video variants undermines the shared community experience of online content and reflects a platform driven by metrics rather than creative conviction.
YouTube has announced new AI-powered features in its YouTube Studio app, including AI feedback on drafts, dynamic thumbnails, and enhanced analytics, to help creators analyze videos and grow their audiences.
This paper proposes A/B Agent, a closed-loop agent framework that organizes historical A/B testing knowledge into a hierarchical experience tree, retrieves transferable strategies via multi-path Tree-RAG, and self-evolves through online experiment feedback, achieving a 4.829% GMV improvement in a short-video e-commerce recommendation system.
Researchers present a practical variance reduction framework combining post-stratification with CUPED for heavy-tailed monetization metrics in ranking experiments, deployed at ShareChat to achieve equivalent statistical confidence with 45% less traffic. The paper is accepted at SIGIR 2026.
SimGym is a framework that simulates A/B tests on e-commerce storefronts using vision-language model agents, reducing experimental cycles from weeks to under an hour while achieving 77% directional alignment with real buyer behavior.