NLGR: Utilizing Neighbor Lists for Generative Rerank in Personalized Recommendation Systems

TL;DR

This paper tackles the goal inconsistency problem in generative reranking by introducing NLGR, which uses neighbor lists to provide relative guidance in the combinatorial space and a sampling-based non-autoregressive generator to explore nearby permutations. The evaluator NLGR-E guides offline training, while the generator NLGR-G leverages Position Decision Unit and Candidate Retrieval Unit to efficiently replace items, guided by counterfactual rewards derived from NLGR-E. Extensive offline experiments on Taobao Ad and Meituan datasets, plus an online A/B test on Meituan, demonstrate significant improvements over strong baselines in AUC, NDCG, HR, CTR, and GMV, justifying deployment in production. Overall, NLGR achieves superior list-wise optimization by aligning generator guidance with evaluator feedback through neighbor-list training and flexible, non-autoregressive generation, delivering practical gains in industrial recommender systems.

Abstract

Reranking plays a crucial role in modern multi-stage recommender systems by rearranging the initial ranking list. Due to the inherent challenges of combinatorial search spaces, some current research adopts an evaluator-generator paradigm, with a generator generating feasible sequences and an evaluator selecting the best sequence based on the estimated list utility. However, these methods still face two issues. Firstly, due to the goal inconsistency problem between the evaluator and generator, the generator tends to fit the local optimal solution of exposure distribution rather than combinatorial space optimization. Secondly, the strategy of generating target items one by one is difficult to achieve optimality because it ignores the information of subsequent items. To address these issues, we propose a utilizing Neighbor Lists model for Generative Reranking (NLGR), which aims to improve the performance of the generator in the combinatorial space. NLGR follows the evaluator-generator paradigm and improves the generator's training and generating methods. Specifically, we use neighbor lists in combination space to enhance the training process, making the generator perceive the relative scores and find the optimization direction. Furthermore, we propose a novel sampling-based non-autoregressive generation method, which allows the generator to jump flexibly from the current list to any neighbor list. Extensive experiments on public and industrial datasets validate NLGR's effectiveness and we have successfully deployed NLGR on the Meituan food delivery platform.

Figures (4)

• Figure 1: Generator optimization legend: reach the optimal step by step under the guidance of the neighbor lists.
• Figure 2: The overall architecture of NLGR.
• Figure 3: The training process of NLGR-G for a candidate list of length 3.
• Figure 4: Architecture of the online deployment with NLGR.