Every Preference Has Its Strength: Injecting Ordinal Semantics into LLM-Based Recommenders

TL;DR

This paper introduces Ordinal Semantic Anchoring (OSA), a novel framework that explicitly models preference strength in LLM-based recommenders by representing ordinal levels as numeric tokens, improving fine-grained preference understanding.

Contribution

OSA is the first method to incorporate explicit ordinal preference levels into LLM-based recommenders, enhancing their ability to capture nuanced user preferences.

Findings

OSA outperforms existing CF-LLM baselines on multiple datasets.

It improves pairwise preference evaluation accuracy.

Explicit modeling of preference strength enhances recommendation quality.

Abstract

Recent work has shown that large language models (LLMs) can enhance recommender systems by integrating collaborative filtering (CF) signals through hybrid prompting. However, most existing CF-LLM frameworks collapse explicit ratings into implicit or positive-only feedback, discarding the ordinal structure that conveys fine-grained preference strength. As a result, these models struggle to exploit graded semantics and nuanced preference distinctions. We propose Ordinal Semantic Anchoring (OSA), a hybrid CF-LLM framework that explicitly incorporates preference strength by modeling interaction-level user feedback. OSA represents ordinal preference levels as numeric textual tokens and uses their token embeddings as semantic anchors to align user-item interaction representations in the LLM latent space. Through strength-aware alignment across ordinal levels, OSA preserves preference semantics when integrating collaborative signals with LLMs. Experiments on multiple real-world datasets demonstrate that OSA consistently outperforms existing baselines, particularly in pairwise preference evaluation, highlighting its effectiveness in modeling fine-grained user preferences over prior CF-LLM methods.