PTCBENCH: Benchmarking Contextual Stability of Personality Traits in LLM Systems

Jiongchi Yu ∗, Yuhan Ma ∗, Xiaoyu Zhang †, Junjie Wang, Qiang Hu, Chao Shen, Xiaofei Xie · 2026

Read PTCBENCH: Benchmarking Contextual Stability of Personality Traits in LLM Systems by Jiongchi Yu ∗, Yuhan Ma ∗, Xiaoyu Zhang †, Junjie Wang, Qiang Hu, Chao Shen, Xiaofei Xie (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-00016.txt.

Description

Jiongchi Yu and coauthors build PTCBENCH to test whether language models keep a steady personality across settings. Personality in such models matters for trust, yet traits are known to shift with context, a point prior work often missed. The test puts models through 12 settings, from places to life events, and scores them with the NEO Five Factor Inventory. Across more than 39000 trait records, settings such as unemployment shift traits sharply and can even change reasoning. The authors offer the bench as a way to build steadier, more human aligned systems.

Catalog record

CREATORJiongchi Yu ∗, Yuhan Ma ∗, Xiaoyu Zhang †, Junjie Wang, Qiang Hu, Chao Shen, Xiaofei Xie
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE70869
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.00016

Browse the full library · About the Tower of Babel