LettinGo: Explore User Profile Generation for Recommendation System

2025-06-23

AI Summary

Existing user profiling methods rely on fixed, rigid structures, limiting their ability to capture behavioral diversity and lacking adaptability and interpretability. To address these limitations, we propose a three-stage dynamic profiling framework: (1) leveraging collaborative large language models (LLMs) to generate semantically rich, stylistically diverse textual user profiles; (2) automatically constructing preference data grounded in recommendation task performance; and (3) applying unsupervised Direct Preference Optimization (DPO) for end-to-end alignment between generated profiles and downstream recommendation objectives—bypassing conventional supervised fine-tuning (SFT) and rigid output-format constraints. Our approach significantly enhances profile expressiveness, personalization capability, and contextual awareness. Empirical evaluation across multiple recommendation benchmarks demonstrates consistent superiority over state-of-the-art profiling methods. The framework is flexible, practically deployable, and inherently interpretable due to its text-based, preference-driven optimization.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Knowledge Representation and Reasoning: Preferences

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

User profiling is pivotal for recommendation systems, as it transforms raw user interaction data into concise and structured representations that drive personalized recommendations. While traditional embedding-based profiles lack interpretability and adaptability, recent advances with large language models (LLMs) enable text-based profiles that are semantically richer and more transparent. However, existing methods often adhere to fixed formats that limit their ability to capture the full diversity of user behaviors. In this paper, we introduce LettinGo, a novel framework for generating diverse and adaptive user profiles. By leveraging the expressive power of LLMs and incorporating direct feedback from downstream recommendation tasks, our approach avoids the rigid constraints imposed by supervised fine-tuning (SFT). Instead, we employ Direct Preference Optimization (DPO) to align the profile generator with task-specific performance, ensuring that the profiles remain adaptive and effective. LettinGo operates in three stages: (1) exploring diverse user profiles via multiple LLMs, (2) evaluating profile quality based on their impact in recommendation systems, and (3) aligning the profile generation through pairwise preference data derived from task performance. Experimental results demonstrate that our framework significantly enhances recommendation accuracy, flexibility, and contextual awareness. This work enhances profile generation as a key innovation for next-generation recommendation systems.

Problem

• Enhances interpretability and adaptability of user profiles
• Overcomes fixed-format limitations in capturing user diversity
• Improves recommendation accuracy via adaptive profile generation

Innovation

• Leverages LLMs for diverse user profiles
• Uses DPO for task-specific profile alignment
• Three-stage framework enhances recommendation accuracy