Token-Controlled Re-ranking for Sequential Recommendation via LLMs

TL;DR

This paper tackles the lack of fine-grained user control in LLM-based sequential recommender systems by introducing COREC, a token-augmented re-ranking framework that injects explicit attribute-level control signals via control tokens. It combines a lightweight item retriever (SASRec) with a token-driven input construction and a RankNet-style fine-tuning objective, producing rankings that respect user-specified constraints while preserving personalization. Empirical results on Amazon 2018 subsets show substantial gains in both standard ranking metrics and novel controllability metrics (CP and CD) compared with strong baselines, and analyses reveal the importance of token-based control and calibrated thresholds over hard filtering. The work advances interactive, controllable recommender systems and lays groundwork for broader multi-attribute, session-based constraint control in practical deployments.

Abstract

The widespread adoption of Large Language Models (LLMs) as re-rankers is shifting recommender systems towards a user-centric paradigm. However, a significant gap remains: current re-rankers often lack mechanisms for fine-grained user control. They struggle to balance inherent user preferences with multiple attribute-based constraints, often resorting to simplistic hard filtering that can excessively narrow the recommendation pool and yield suboptimal results. This limitation leaves users as passive recipients rather than active collaborators in the recommendation process. To bridge this gap, we propose COREC, a novel token-augmented re-ranking framework that incorporates specific user requirements in co-creating the recommendation outcome. COREC empowers users to steer re-ranking results with precise and flexible control via explicit, attribute-based signals. The framework learns to balance these commands against latent preferences, yielding rankings that adhere to user instructions without sacrificing personalization. Experiments show that COREC: (1) exceeds state-of-the-art baselines on standard recommendation effectiveness and (2) demonstrates superior adherence to specific attribute requirements, proving that COREC enables fine-grained and predictable manipulation of the rankings.

Figures (12)

• Figure 1: (a) Previous methods in recommender systems have largely focused on inferring potential preferences. This paradigm places users in a passive position. (b) In contrast, CoRec allows users to regulate recommendation results according to fine-grained attributes.
• Figure 2: Illustration of our input. We use blue to mark components aligned with the control scheme: the scheme itself and matching control tokens. Gray is used for non-matching token elements and uncontrolled metadata.
• Figure 3: Pipeline of CoRec. In the retrieval stage, we generate candidate items for each window. During the LLM re-ranking stage, the input sequence is constructed, injecting control tokens, and the LLM is then trained based on the ranking module.
• Figure 4: (a) Distribution of <price> in Home training set. (b) Distribution of <category> in Beauty training set.
• Figure 5: Zero-shot vs CoRec comparison on Home dataset. (a) Two-token scheme; (b) Three-token scheme.