Sparse Personalized Text Generation with Multi-Trajectory Reasoning

2026-04-27

AI Summary

This work addresses the challenge of effective personalized text generation with large language models in cold-start scenarios where user interaction data is sparse or absent. To this end, the authors propose the PAT framework, which introduces a novel multi-trajectory alignment inference mechanism. PAT retrieves writing styles from style-similar users and contextual topics from preference-aligned users, then fuses these heterogeneous signals through a reinforcement learning–driven iterative dual-inference process. By integrating retrieval augmentation with fine-tuning of large language models, PAT significantly improves both generation quality and user alignment across multiple real-world personalized benchmarks. The approach effectively overcomes the reliance on dense historical interaction data inherent in conventional methods, offering a robust solution for cold-start personalized generation.

Abstract

As Large Language Models (LLMs) advance, personalization has become a key mechanism for tailoring outputs to individual user needs. However, most existing methods rely heavily on dense interaction histories, making them ineffective in cold-start scenarios where such data is sparse or unavailable. While external signals (e.g., content of similar users) can offer a potential remedy, leveraging them effectively remains challenging: raw context is often noisy, and existing methods struggle to reason over heterogeneous data sources. To address these issues, we introduce PAT (Personalization with Aligned Trajectories), a reasoning framework for cold-start LLM personalization. PAT first retrieves information along two complementary trajectories: writing-style cues from stylistically similar users and topic-specific context from preference-aligned users. It then employs a reinforcement learning-based, iterative dual-reasoning mechanism that enables the LLM to jointly refine and integrate these signals. Experimental results across real-world personalization benchmarks show that PAT consistently improves generation quality and alignment under sparse-data conditions, establishing a strong solution to the cold-start personalization problem.

Problem

• cold-start personalization
• sparse interaction data
• heterogeneous data sources
• personalized text generation
• large language models

Innovation

• cold-start personalization
• multi-trajectory reasoning
• sparse data
• reinforcement learning
• large language models