Autoregressive Generation Strategies for Top-K Sequential Recommendations

The goal of modern sequential recommender systems is often formulated in terms of next-item prediction. In this paper, we explore the applicability of transformerbased generative models for the Top-K sequential recommendation task, where the goal is to predict items that a user is likely to interact with in the “near future”. This goal aligns with real-world applications of such models in an offline scenario or as a part of multi-stage recommender pipelines. We explore commonly used autoregressive generation strategies, including greedy decoding, beam search, and temperature sampling, to evaluate their performance for the Top-K sequential recommendation task. In addition, we propose novel Reciprocal Rank Aggregation (RRA) and Relevance Aggregation (RA) generation strategies based on multi-sequence generation with temperature sampling and subsequent aggregation. Experiments on diverse datasets give valuable insights regarding the applicability of commonly used strategies and show that the suggested approaches improve performance on longer time horizons compared to the widely used Top-K prediction approach and single-sequence autoregressive generation strategies.