Emergent Hierarchical Structure in Large Language Models: An Information-Theoretic Framework for Multi-Scale Representation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zhang, Yukin, Dong, Qi, Xu, Kemu
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913091638263808
author Zhang, Yukin
Dong, Qi
Xu, Kemu
author_facet Zhang, Yukin
Dong, Qi
Xu, Kemu
contents Why do language models from different architecture families respond so differently to the same perturbation? We argue that the answer is not scale, but \emph{how architecture shapes information compression}. Analyzing eight Transformer models (7B--70B parameters) from the Llama and Qwen families, we show that every model spontaneously develops discrete functional boundaries dividing its layers into Local, Intermediate, and Global processing segments -- yet boundary locations and per-segment brittleness are determined overwhelmingly by architecture family rather than model size or training configuration. We formalize this regularity as the \textbf{Multi-Scale Probabilistic Generation Theory} (MSPGT), which models an autoregressive Transformer as a Hierarchical Variational Information Bottleneck system and derives a tiered set of falsifiable predictions. Three predictions are strongly confirmed: all eight models exhibit two prominent phase-transition boundaries (P1.1); Llama boundary positions are stable across a $10{\times}$ parameter range ($\mathrm{CV}{=}0.067$--$0.095$) while Qwen positions vary widely ($\mathrm{CV}{=}0.465$--$0.726$), precisely matching our strong- and weak-dominance conditions; and cross-architecture local-segment brittleness spans \textbf{three orders of magnitude} ($493{\times}$ ratio) -- a gap that architecture family alone predicts and that dwarfs any within-family or scale-driven variation.
format Preprint
id arxiv_https___arxiv_org_abs_2505_18244
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Emergent Hierarchical Structure in Large Language Models: An Information-Theoretic Framework for Multi-Scale Representation
Zhang, Yukin
Dong, Qi
Xu, Kemu
Computation and Language
Artificial Intelligence
Why do language models from different architecture families respond so differently to the same perturbation? We argue that the answer is not scale, but \emph{how architecture shapes information compression}. Analyzing eight Transformer models (7B--70B parameters) from the Llama and Qwen families, we show that every model spontaneously develops discrete functional boundaries dividing its layers into Local, Intermediate, and Global processing segments -- yet boundary locations and per-segment brittleness are determined overwhelmingly by architecture family rather than model size or training configuration. We formalize this regularity as the \textbf{Multi-Scale Probabilistic Generation Theory} (MSPGT), which models an autoregressive Transformer as a Hierarchical Variational Information Bottleneck system and derives a tiered set of falsifiable predictions. Three predictions are strongly confirmed: all eight models exhibit two prominent phase-transition boundaries (P1.1); Llama boundary positions are stable across a $10{\times}$ parameter range ($\mathrm{CV}{=}0.067$--$0.095$) while Qwen positions vary widely ($\mathrm{CV}{=}0.465$--$0.726$), precisely matching our strong- and weak-dominance conditions; and cross-architecture local-segment brittleness spans \textbf{three orders of magnitude} ($493{\times}$ ratio) -- a gap that architecture family alone predicts and that dwarfs any within-family or scale-driven variation.
title Emergent Hierarchical Structure in Large Language Models: An Information-Theoretic Framework for Multi-Scale Representation
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2505.18244