arXiv 2509.04504
Behavioral Fingerprinting of Large Language Models
By Zehua Pei, Hui-Ling Zhen, et al.
Published 2025-09-02
Mindmap
Browse the paper's core ideas, clusters, and relationships in a structured outline.
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 Diagnostic P…