Behavioral Fingerprinting of Large Language Models

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Pei, Zehua, Zhen, Hui-Ling, Zhang, Ying, Yang, Zhiyuan, Li, Xing, Yu, Xianzhi, Yuan, Mingxuan, Yu, Bei
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911138755641344
author Pei, Zehua
Zhen, Hui-Ling
Zhang, Ying
Yang, Zhiyuan
Li, Xing
Yu, Xianzhi
Yuan, Mingxuan
Yu, Bei
author_facet Pei, Zehua
Zhen, Hui-Ling
Zhang, Ying
Yang, Zhiyuan
Li, Xing
Yu, Xianzhi
Yuan, Mingxuan
Yu, Bei
contents Current benchmarks for Large Language Models (LLMs) primarily focus on performance metrics, often failing to capture the nuanced behavioral characteristics that differentiate them. This paper introduces a novel ``Behavioral Fingerprinting'' framework designed to move beyond traditional evaluation by creating a multi-faceted profile of a model's intrinsic cognitive and interactive styles. Using a curated \textit{Diagnostic Prompt Suite} and an innovative, automated evaluation pipeline where a powerful LLM acts as an impartial judge, we analyze eighteen models across capability tiers. Our results reveal a critical divergence in the LLM landscape: while core capabilities like abstract and causal reasoning are converging among top models, alignment-related behaviors such as sycophancy and semantic robustness vary dramatically. We further document a cross-model default persona clustering (ISTJ/ESTJ) that likely reflects common alignment incentives. Taken together, this suggests that a model's interactive nature is not an emergent property of its scale or reasoning power, but a direct consequence of specific, and highly variable, developer alignment strategies. Our framework provides a reproducible and scalable methodology for uncovering these deep behavioral differences. Project: https://github.com/JarvisPei/Behavioral-Fingerprinting
format Preprint
id arxiv_https___arxiv_org_abs_2509_04504
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Behavioral Fingerprinting of Large Language Models
Pei, Zehua
Zhen, Hui-Ling
Zhang, Ying
Yang, Zhiyuan
Li, Xing
Yu, Xianzhi
Yuan, Mingxuan
Yu, Bei
Computation and Language
Artificial Intelligence
Current benchmarks for Large Language Models (LLMs) primarily focus on performance metrics, often failing to capture the nuanced behavioral characteristics that differentiate them. This paper introduces a novel ``Behavioral Fingerprinting'' framework designed to move beyond traditional evaluation by creating a multi-faceted profile of a model's intrinsic cognitive and interactive styles. Using a curated \textit{Diagnostic Prompt Suite} and an innovative, automated evaluation pipeline where a powerful LLM acts as an impartial judge, we analyze eighteen models across capability tiers. Our results reveal a critical divergence in the LLM landscape: while core capabilities like abstract and causal reasoning are converging among top models, alignment-related behaviors such as sycophancy and semantic robustness vary dramatically. We further document a cross-model default persona clustering (ISTJ/ESTJ) that likely reflects common alignment incentives. Taken together, this suggests that a model's interactive nature is not an emergent property of its scale or reasoning power, but a direct consequence of specific, and highly variable, developer alignment strategies. Our framework provides a reproducible and scalable methodology for uncovering these deep behavioral differences. Project: https://github.com/JarvisPei/Behavioral-Fingerprinting
title Behavioral Fingerprinting of Large Language Models
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2509.04504