CoFiRec: Coarse-to-Fine Tokenization for Generative Recommendation

2025-11-27

AI Summary

Existing generative recommender systems overlook the user’s intent evolution—from broad categories to specific items—during browsing behavior and flatten heterogeneous item attributes into a single embedding, thereby eroding inherent semantic hierarchies. To address this, we propose CoFiRec, a fine-grained generative recommendation framework that explicitly models item semantics (category → descriptive attributes → collaborative signals) within the tokenization process. CoFiRec introduces multi-level independent tokenizers and an autoregressive decoding mechanism to enable coarse-to-fine, progressive item generation. We theoretically prove that structured tokenization reduces generative bias. Extensive experiments across multiple benchmarks and backbone architectures demonstrate significant improvements over state-of-the-art methods, validating that hierarchical semantic modeling effectively captures dynamic user intent.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders
• Data Mining & Knowledge Management: Recommender Systems

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

In web environments, user preferences are often refined progressively as users move from browsing broad categories to exploring specific items. However, existing generative recommenders overlook this natural refinement process. Generative recommendation formulates next-item prediction as autoregressive generation over tokenized user histories, where each item is represented as a sequence of discrete tokens. Prior models typically fuse heterogeneous attributes such as ID, category, title, and description into a single embedding before quantization, which flattens the inherent semantic hierarchy of items and fails to capture the gradual evolution of user intent during web interactions. To address this limitation, we propose CoFiRec, a novel generative recommendation framework that explicitly incorporates the Coarse-to-Fine nature of item semantics into the tokenization process. Instead of compressing all attributes into a single latent space, CoFiRec decomposes item information into multiple semantic levels, ranging from high-level categories to detailed descriptions and collaborative filtering signals. Based on this design, we introduce the CoFiRec Tokenizer, which tokenizes each level independently while preserving structural order. During autoregressive decoding, the language model is instructed to generate item tokens from coarse to fine, progressively modeling user intent from general interests to specific item-level interests. Experiments across multiple public benchmarks and backbones demonstrate that CoFiRec outperforms existing methods, offering a new perspective for generative recommendation. Theoretically, we prove that structured tokenization leads to lower dissimilarity between generated and ground truth items, supporting its effectiveness in generative recommendation. Our code is available at https://github.com/YennNing/CoFiRec.

Problem

• Captures gradual user intent evolution in web interactions
• Preserves semantic hierarchy in item tokenization for recommendation
• Models user preferences from coarse to fine granularity

Innovation

• Coarse-to-fine tokenization captures semantic hierarchy
• Independent tokenization of multiple semantic levels
• Autoregressive decoding from coarse to fine intent