MultiSlot ReRanker: A Generic Model-based Re-Ranking Framework in Recommendation Systems

TL;DR

This work tackles multislot re-ranking for whole-page recommendations by jointly optimizing relevance, diversity, and freshness. It introduces a Sequential Greedy Algorithm with linear-time complexity and a multislot offline replay mechanism, augmented by Pareto-optimal enhancements and an OpenAI Gym–Ray simulator for rapid benchmarking. Empirical results show offline AUC gains in the range of $+6\%$ to $+10\%$, driven by explicit modeling of mutual influences among items and leveraging multi-objective second-pass scores; the offline replay framework enables evaluation under multiple trade-offs via Pareto Frontiers. The accompanying MultiSlot Simulator supports fast experimentation with RL and supervised methods, facilitating production-ready deployment and future work on multi-step optimization.

Abstract

In this paper, we propose a generic model-based re-ranking framework, MultiSlot ReRanker, which simultaneously optimizes relevance, diversity, and freshness. Specifically, our Sequential Greedy Algorithm (SGA) is efficient enough (linear time complexity) for large-scale production recommendation engines. It achieved a lift of $+6\%$ to $ +10\%$ offline Area Under the receiver operating characteristic Curve (AUC) which is mainly due to explicitly modeling mutual influences among items of a list, and leveraging the second pass ranking scores of multiple objectives. In addition, we have generalized the offline replay theory to multi-slot re-ranking scenarios, with trade-offs among multiple objectives. The offline replay results can be further improved by Pareto Optimality. Moreover, we've built a multi-slot re-ranking simulator based on OpenAI Gym integrated with the Ray framework. It can be easily configured for different assumptions to quickly benchmark both reinforcement learning and supervised learning algorithms.

Figures (3)

• Figure 1: The MultiSlot ReRanker Framework (e.g., with 20 Slots).
• Figure 2: Replay Result of MultiSlot ReRanker XGBoost.
• Figure 3: Replay Result of MultiSlot ReRanker XGBoost with Pareto Optimality.