DeepInterestGR: Mining Deep Multi-Interest Using Multi-Modal LLMs for Generative Recommendation

TL;DR

DeepInterestGR advances generative recommendation by mining deep, multi-modal user interests using large language models and reinforcement learning, significantly improving personalization and interpretability over shallow methods.

Contribution

It introduces a multi-modal interest mining framework, reward-based supervision, and interest-aware item encoding, enabling deeper user interest modeling in generative recommendation systems.

Findings

Outperforms state-of-the-art baselines on Amazon Review benchmarks.

Achieves higher HR@K and NDCG@K metrics.

Effectively captures deep, multi-modal user interests.

Abstract

Recent generative recommendation frameworks have demonstrated remarkable scaling potential by reformulating item prediction as autoregressive Semantic ID (SID) generation. However, existing methods primarily rely on shallow behavioral signals, encoding items solely through surface-level textual features such as titles and descriptions. This reliance results in a critical Shallow Interest problem: the model fails to capture the latent, semantically rich interests underlying user interactions, limiting both personalization depth and recommendation interpretability. DeepInterestGR introduces three key innovations: (1) Multi-LLM Interest Mining (MLIM): We leverage multiple frontier LLMs along with their multi-modal variants to extract deep textual and visual interest representations through Chain-of-Thought prompting. (2) Reward-Labeled Deep Interest (RLDI): We employ a lightweight binary classifier to assign reward labels to mined interests, enabling effective supervision signals for reinforcement learning. (3) Interest-Enhanced Item Discretization (IEID): The curated deep interests are encoded into semantic embeddings and quantized into SID tokens via RQ-VAE. We adopt a two-stage training pipeline: supervised fine-tuning aligns the generative model with deep interest signals and collaborative filtering patterns, followed by reinforcement learning with GRPO optimized by our Interest-Aware Reward. Experiments on three Amazon Review benchmarks demonstrate that DeepInterestGR consistently outperforms state-of-the-art baselines across HR@K and NDCG@K metrics.