HoloRec: Holistic Encoding and Interleaved Reasoning for Generative Recommendation
This work addresses the lack of hierarchical reasoning and expensive external annotations in generative recommendation models, offering a unified framework that improves accuracy, especially in sparse data settings.
HoloRec introduces an endogenous chain-of-thought mechanism for generative recommendation that unifies representation, reasoning, and generation via hierarchical semantic encoding and interleaved reasoning, outperforming baselines with significant gains in sparse scenarios.
Generative recommendation models that formulate the task as sequence generation overcome the objective fragmentation problem of traditional cascade architectures, yet existing approaches still suffer from flat semantic representations lacking hierarchical structure for multi-step reasoning and an externally constructed chain-of-thought (CoT) that requires expensive annotations and remains disconnected from the generation objective. We propose HoloRec, an endogenous chain-of-thought recommendation mechanism that unifies representation, reasoning, and generation by constructing a hierarchical semantic encoding matrix via multi-granularity nested residual quantization optimized by a holistic reconstruction loss. HoloRec supports two inference modes: a non-thinking mode that uses lightweight multi-granularity supervised alignment for fast prediction, and a thinking mode that employs an interleaved reasoning scheme to generate CoT steps on the fly, directly embedding reasoning into the generation process without external data. Experiments on multiple public recommendation datasets demonstrate that HoloRec consistently outperforms baselines, with especially significant gains in sparse scenarios, and the thinking mode achieves better accuracy than the non-thinking mode with only modest inference overhead.