捆綁推薦（Bundle Recommendation）——將多個單品（Item）組合為捆綁包（Bundle）進(jìn)行推薦，而非推薦單個物品——正成為智能推薦領(lǐng)域一個令人興奮的拓展方向。

近五年來，我們團(tuán)隊圍繞這一主題，從數(shù)據(jù)資源、方法復(fù)現(xiàn)、系統(tǒng)綜述到創(chuàng)新方法四個維度持續(xù)深耕，陸續(xù)收獲了一些"小確幸"：

1. 發(fā)布食品領(lǐng)域首個兼顧個性化與健康性的套餐推薦數(shù)據(jù)集, MealRec+: A Meal Recommendation Dataset with Meal-Course Affiliation for Personalization and Healthiness, SIGIR 2024.

https://dl.acm.org/doi/10.1145/3626772.3657857

2. 探索"負(fù)責(zé)任的推薦"——讓算法既懂你的口味，也關(guān)心你的健康，Boosting Healthiness Exposure in Category-Constrained Meal Recommendation Using Nutritional Standards，ACM TIST 2024.

https://dl.acm.org/doi/10.1145/3643859

3. 從捆綁編輯到捆綁推薦，系統(tǒng)復(fù)現(xiàn)并剖析主流方法，A Reproducibility Study of Bundle Editing and Bundle Recommendation, SIGIR 2026.

https://dl.acm.org/doi/10.1145/3805712.3808559

4. 全面綜述判別式和生成式捆綁推薦的研究進(jìn)展，A Survey on Bundle Recommendation: Methods, Applications, and Challenges，ACM Computing Surveys, 2026.

https://dl.acm.org/doi/10.1145/3802820

5. 發(fā)現(xiàn)捆綁冷啟動具有冷暖混合的特性, Divide-and-Conquer: Cold-Start Bundle Recommendation via Mixture of Diffusion Experts, ACM TOIS, 2026.

https://dl.acm.org/doi/10.1145/3799250

從一個看似"小眾"的課題出發(fā)，卻在深入探索中不斷邂逅有趣的發(fā)現(xiàn)——這大概就是興趣驅(qū)動型科研的魅力所在吧 ?