LaRec: Unleashing LLM-based Latent Reasoning for Generative Recommendation
2026-07-27 • Information Retrieval
Information Retrieval
AI summaryⓘ
The authors propose LaRec, a recommendation system that improves how large language models think behind the scenes by using a smarter, hidden way of reasoning called latent reasoning. They address two problems: that the model doesn't get enough detailed guidance while learning and that it usually follows only one way of thinking, which limits understanding user preferences. LaRec first trains the model to better handle hidden reasoning steps and then uses a personalized method to explore different interests by sampling from a mix of user preferences. Their tests show LaRec works better than previous methods without being slower.
Large Language ModelsRecommendation SystemsLatent ReasoningChain-of-ThoughtSupervised LearningReinforcement LearningGaussian Mixture ModelPersonalizationLatent SpaceUser Preferences
Authors
Yu Xia, Zihan Lin, Wei Yang, Rui Zhong, Cheng Chen, Huan Ren, Yao Hu
Abstract
Large Language Models (LLMs) have shown great promise in recommendation due to superior reasoning abilities. However, existing methods mainly rely on explicit Chain-of-Thought (CoT), resulting in verbose reasoning texts and inefficient response times. latent reasoning aims to balance efficiency by thinking within a continuous latent space, yet it faces two major challenges: (1) Lack of Fine-grained Supervision: Latent reasoning relies solely on feedback from the final labels, providing sparse supervisory signals that struggle to effectively guide the optimization of multiple hidden reasoning steps. (2) Single Reasoning Path: The deterministic nature of latent reasoning impedes the exploration of users' diverse interests and preferences, thereby limiting the recommendation capabilities of LLMs. To address these issues, we propose \textbf{$LaRec$}, an efficient generative recommendation framework designed to unleash the potential of latent reasoning in LLMs. $LaRec$ consists of two core stages: First, we design Latent Pre-training that empowers LLMs with latent reasoning capabilities by providing rich supervisory signals to the latent space reasoning via step-level alignment and process direction alignment. Second, we introduce Personalized RL-tuning. Specifically, we construct a personalized Gaussian Mixture Distribution for each user based on their historical interests. By randomly sampling distinct reasoning starting points from this distribution during training, we guide the LLMs to traverse diverse reasoning paths within the latent space, enabling efficient exploration of user's multi-faceted interests. Experiments on multiple datasets show that $LaRec$ significantly outperforms existing baselines with comparable efficiency.