重审柏拉图式表征假说：一种亚里士多德式的视角

Revisiting the Platonic Representation Hypothesis: An Aristotelian View

February 16, 2026

作者: Fabian Gröger, Shuo Wen, Maria Brbić

cs.AI

摘要

柏拉图式表征假说认为，神经网络的表征正在趋同于现实的统一统计模型。我们发现，现有衡量表征相似度的指标受到网络规模的干扰：增加模型深度或宽度会系统性地抬高表征相似度评分。为修正这些影响，我们提出基于排列的零校准框架，可将任何表征相似度指标转化为具有统计保证的校准分数。通过我们的校准框架重新审视柏拉图式表征假说，发现了一个微妙图景：全局谱度量所报告的明显趋同现象在校准后基本消失，而局部邻域相似性（非局部距离）在不同模态间仍保持显著一致性。基于这些发现，我们提出亚里士多德式表征假说：神经网络表征正在趋同于共享的局部邻域关系。

English

The Platonic Representation Hypothesis suggests that representations from neural networks are converging to a common statistical model of reality. We show that the existing metrics used to measure representational similarity are confounded by network scale: increasing model depth or width can systematically inflate representational similarity scores. To correct these effects, we introduce a permutation-based null-calibration framework that transforms any representational similarity metric into a calibrated score with statistical guarantees. We revisit the Platonic Representation Hypothesis with our calibration framework, which reveals a nuanced picture: the apparent convergence reported by global spectral measures largely disappears after calibration, while local neighborhood similarity, but not local distances, retains significant agreement across different modalities. Based on these findings, we propose the Aristotelian Representation Hypothesis: representations in neural networks are converging to shared local neighborhood relationships.