Benchmarking and Improving LLM Robustness for Personalized Generation

2025-09-18

AI Summary

This paper addresses the imbalance between factual accuracy and user preference alignment in personalized generation by large language models (LLMs). We first systematically define and quantify “personalized generation robustness,” revealing a critical factuality–preference trade-off: e.g., GPT-4.1 exhibits a 5% factual error rate, while smaller models exceed 20%. To tackle this, we propose PERG—a unified evaluation framework—and its benchmark dataset PERGData, alongside Pref-Aligner, a two-stage optimization method for jointly improving factual consistency and preference fidelity. Through multi-model comparisons, diverse prompting strategies, and human-in-the-loop evaluation, we uncover how query types and preference categories modulate robustness. Experiments demonstrate that Pref-Aligner improves average robustness by 25% across mainstream LLMs, establishing a new benchmark and scalable paradigm for trustworthy, preference-aware generation.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Learning Preferences or Rankings
• Humans and AI: Learning Human Values and Preferences

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Recent years have witnessed a growing interest in personalizing the responses of large language models (LLMs). While existing evaluations primarily focus on whether a response aligns with a user's preferences, we argue that factuality is an equally important yet often overlooked dimension. In the context of personalization, we define a model as robust if its responses are both factually accurate and align with the user preferences. To assess this, we introduce PERG, a scalable framework for evaluating robustness in LLMs, along with a new dataset, PERGData. We evaluate fourteen models from five different model families using different prompting methods. Our findings show that current LLMs struggle with robust personalization: even the strongest models (GPT-4.1, LLaMA3-70B) fail to maintain correctness in 5% of previously successful cases without personalization, while smaller models (e.g., 7B-scale) can fail more than 20% of the time. Further analysis reveals that robustness is significantly affected by the nature of the query and the type of user preference. To mitigate these failures, we propose Pref-Aligner, a two-stage approach that improves robustness by an average of 25% across models. Our work highlights critical gaps in current evaluation practices and introduces tools and metrics to support more reliable, user-aligned LLM deployments.

Problem

• Evaluating LLM robustness for personalized generation balancing factuality and user preferences
• Addressing significant failure rates in maintaining correctness during personalized responses
• Improving robustness against query nature and user preference type variations

Innovation

• PERG framework for evaluating LLM robustness
• Pref-Aligner two-stage approach for improvement
• Factuality and preference alignment as robustness metrics