软课程学习优化新鲜与泛化推荐
原标题：Soft Curriculum Learning for Optimizing Fresh and Generalized Recommendations
AI 摘要
该论文提出一种可扩展的软课程学习（Soft Curriculum Learning）框架，用于工业级推荐系统的持续训练场景。针对短视频平台中流行度反馈循环导致模型偏向头部内容、忽视长尾内容的问题，作者通过损失退火和图内权重调整替代刚性数据过滤，在不牺牲吞吐量的前提下打破流行度反馈。方法在SASRec、双塔检索模型和大规模持续训练排序模型上验证，在线A/B测试显示用户满意度和新鲜内容消费均有显著提升。
正文节选
Soft Curriculum Learning for Optimizing Fresh and Generalized RecommendationsCCS: Information systems PersonalizationCCS: Information systems Recommender systemsConference: 20th ACM Conference on Recommender Systems; September 27-October 02, 2026; Minneapolis, MN, USA20th ACM Conference on Recommender Systems (RecSys ’26), September 27-October 02, 2026, Minneapolis, MN, USADOI: 10.1145/3773078.3831923ISBN: 979-8-4007-2284-4/2026/09 Abstract. Large-scale recommender systems, particularly short-fo