0:000:00
<p>he research introduces <strong>OneRec</strong>, a novel generative recommender system designed to unify the traditional multi-stage "retrieve-and-rank" process into a single, end-to-end generative model. This unified approach, implemented with an <strong>encoder-decoder architecture</strong> and a <strong>sparse Mixture-of-Experts (MoE)</strong> structure for scalable capacity, overcomes the limitations of cascaded ranking systems. Crucially, OneRec employs a <strong>session-wise generation</strong> method, predicting a list of coherent items rather than just the next item, and incorporates an <strong>Iterative Preference Alignment (IPA)</strong> module using Direct Preference Optimization (DPO) tailored for recommendation sparsity to significantly enhance result quality. The model has been successfully deployed on the Kuaishou platform, demonstrating superior performance by achieving a substantial increase in watch-time metrics.</p><p><a href="https://arxiv.org/pdf/2502.18965">arxi...