Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Peng, Luzhou, Yang, Zhengxin, Ji, Honglu, Yang, Yikang, Fan, Fanda, Gao, Wanling, Ge, Jiayuan, Han, Yilin, Zhan, Jianfeng
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2603.04408
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912943029878784
author Peng, Luzhou
Yang, Zhengxin
Ji, Honglu
Yang, Yikang
Fan, Fanda
Gao, Wanling
Ge, Jiayuan
Han, Yilin
Zhan, Jianfeng
author_facet Peng, Luzhou
Yang, Zhengxin
Ji, Honglu
Yang, Yikang
Fan, Fanda
Gao, Wanling
Ge, Jiayuan
Han, Yilin
Zhan, Jianfeng
contents Current evaluation paradigms for large language models (LLMs) characterize models and datasets separately, yielding coarse descriptions: items in datasets are treated as pre-labeled entries, and models are summarized by overall scores such as accuracy, together ignoring the diversity of population-level model behaviors across items with varying properties. To address this gap, this paper conceptualizes LLMs as composed of memes, a notion introduced by Dawkins as cultural genes that replicate knowledge and behavior. Building on this perspective, the Probing Memes paradigm reconceptualizes evaluation as an entangled world of models and data. It centers on a Perception Matrix that captures model-item interactions, enabling Probe Properties for characterizing items and Meme Scores for depicting model behavioral traits. Applied to 9 datasets and 4,507 LLMs, Probing Memes reveals hidden capability structures and quantifies phenomena invisible under traditional paradigms (e.g., elite models failing on problems that most models answer easily). It not only supports more informative and extensible benchmarks but also enables population-based evaluation of LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2603_04408
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Probing Memes in LLMs: A Paradigm for the Entangled Evaluation World
Peng, Luzhou
Yang, Zhengxin
Ji, Honglu
Yang, Yikang
Fan, Fanda
Gao, Wanling
Ge, Jiayuan
Han, Yilin
Zhan, Jianfeng
Computation and Language
Current evaluation paradigms for large language models (LLMs) characterize models and datasets separately, yielding coarse descriptions: items in datasets are treated as pre-labeled entries, and models are summarized by overall scores such as accuracy, together ignoring the diversity of population-level model behaviors across items with varying properties. To address this gap, this paper conceptualizes LLMs as composed of memes, a notion introduced by Dawkins as cultural genes that replicate knowledge and behavior. Building on this perspective, the Probing Memes paradigm reconceptualizes evaluation as an entangled world of models and data. It centers on a Perception Matrix that captures model-item interactions, enabling Probe Properties for characterizing items and Meme Scores for depicting model behavioral traits. Applied to 9 datasets and 4,507 LLMs, Probing Memes reveals hidden capability structures and quantifies phenomena invisible under traditional paradigms (e.g., elite models failing on problems that most models answer easily). It not only supports more informative and extensible benchmarks but also enables population-based evaluation of LLMs.
title Probing Memes in LLMs: A Paradigm for the Entangled Evaluation World
topic Computation and Language
url https://arxiv.org/abs/2603.04408