Hierarchical Latent Reasoning for LLM-based Recommendation

2026-07-30

AI Summary

This work addresses the limitations of existing large language model (LLM)-based recommendation approaches, which either incur high computational overhead due to explicit natural language reasoning or struggle to effectively model the contribution of multi-layer hidden states to user preferences in implicit reasoning. To overcome these challenges, we propose HiLaR, a novel framework that aligns LLM’s multi-layer hidden states with a temporally guided hierarchical user preference representation, structuring the reasoning process from coarse-grained preferences to fine-grained intents. HiLaR further introduces, for the first time, a layer-aware process reward mechanism that jointly optimizes the reasoning trajectory with the ultimate recommendation objective. Extensive experiments on four Amazon benchmark datasets demonstrate that HiLaR significantly outperforms state-of-the-art sequential, generative, and LLM-based recommendation methods, while ablation studies confirm the effectiveness of its key components.

Abstract

Large Language Models (LLMs) have shown strong potential for recommendation by leveraging their semantic understanding and contextual modeling capabilities. Recent studies further introduce reasoning mechanisms to improve user preference modeling. However, explicit natural-language reasoning incurs substantial inference overhead, whereas existing latent reasoning methods mainly focus on generating or verifying intermediate states, leaving their layer-wise preference roles and contributions insufficiently characterized. We propose HiLaR, a Hierarchical Latent Reasoning framework with layer-aware reinforcement optimization for LLM-based recommendation. HiLaR constructs temporal-guided hierarchical user preference representations, aligns them with multiple LLM latent reasoning states, and organizes the reasoning process from broad preferences to fine-grained current intents. To further optimize the reasoning trajectory, HiLaR combines final recommendation feedback with layer-aware process rewards derived from the marginal target-likelihood gain of each state. Experiments on four Amazon benchmark datasets show that HiLaR generally outperforms strong sequential, generative, and LLM-based recommendation baselines. Ablation and sensitivity analyses further verify the contribution of hierarchical representation learning, latent alignment, and process-level optimization. Our code is available in https://github.com/hupeiyu21/HiLaR.

Problem

• LLM-based recommendation
• latent reasoning
• user preference modeling
• inference overhead
• hierarchical representation

Innovation

• Hierarchical Latent Reasoning
• Layer-aware Optimization
• LLM-based Recommendation
• Preference Representation
• Reinforcement Learning