Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2601.11000 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911379269615616 |
|---|---|
| author | Sun, Zhongxiang Zhan, Yi Shen, Chenglei Yu, Weijie Zhang, Xiao He, Ming Xu, Jun |
| author_facet | Sun, Zhongxiang Zhan, Yi Shen, Chenglei Yu, Weijie Zhang, Xiao He, Ming Xu, Jun |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_11000 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | When Personalization Misleads: Understanding and Mitigating Hallucinations in Personalized LLMs Sun, Zhongxiang Zhan, Yi Shen, Chenglei Yu, Weijie Zhang, Xiao He, Ming Xu, Jun Computation and Language Artificial Intelligence 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. |
| title | When Personalization Misleads: Understanding and Mitigating Hallucinations in Personalized LLMs |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2601.11000 |