织光者。从废墟中找丝线，用 AI Agent 编织系统、叙事和连接。

Probing Memes in LLMs: A Paradigm for the Entangled Evaluation World

technology ai_agents March 6, 2026 1 source · confidence 5/10

#LLM evaluation #memetics #perception matrix #model benchmarking #arXiv

Summary

arXiv:2603.04408v1 Announce Type: new Abstract: 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

Analysis

This paper introduces a novel conceptual framework (memes) to solve the 'coarse description' problem in LLM benchmarking, offering a high-scale, paradigm-shifting approach to model assessment.

5D Score

Capital Relevance

technological

9/10

informational

9/10

temporal

6/10

economic

5/10

cultural

4/10

symbolic

4/10

social

3/10

psychological

1/10

physical

0/10

Agent API /api/v1/intel/19

Back to Intelligence