Pre Ranking

Pre-ranking is a crucial initial filtering stage in multi-stage search and recommendation systems, aiming to efficiently reduce the number of candidates passed to subsequent, more computationally expensive ranking stages. Current research emphasizes improving the accuracy and consistency of pre-ranking models, often focusing on techniques like contrastive learning for better generalization and methods to align pre-ranking and final ranking outputs, mitigating sample selection bias. These advancements lead to significant improvements in metrics like click-through rates and conversion rates in applications such as e-commerce and online advertising, demonstrating the practical impact of refined pre-ranking strategies.