When Personalization Misleads: Understanding and Mitigating Hallucinations in Personalized LLMs
This addresses a critical issue for users of personalized AI systems, where factual reliability is compromised, and is incremental as it builds on existing personalization methods.
The paper tackles the problem of personalization-induced hallucinations in personalized large language models, where models generate answers aligned with user history rather than objective truth, and proposes Factuality-Preserving Personalized Steering (FPPS) to mitigate this, showing substantial improvements in factual accuracy while preserving personalized performance.
Personalized large language models (LLMs) adapt model behavior to individual users to enhance user satisfaction, yet personalization can inadvertently distort factual reasoning. We show that when personalized LLMs face factual queries, there exists a phenomenon where the model generates answers aligned with a user's prior history rather than the objective truth, resulting in personalization-induced hallucinations that degrade factual reliability and may propagate incorrect beliefs, due to representational entanglement between personalization and factual representations. To address this issue, we propose Factuality-Preserving Personalized Steering (FPPS), a lightweight inference-time approach that mitigates personalization-induced factual distortions while preserving personalized behavior. We further introduce PFQABench, the first benchmark designed to jointly evaluate factual and personalized question answering under personalization. Experiments across multiple LLM backbones and personalization methods show that FPPS substantially improves factual accuracy while maintaining personalized performance.