The Lattice Representation Hypothesis of Large Language Models

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteur principal: Xiong, Bo
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866916019515162624
author Xiong, Bo
author_facet Xiong, Bo
contents We propose the Lattice Representation Hypothesis of large language models: a symbolic backbone that grounds conceptual hierarchies and logical operations in embedding geometry. Our framework unifies the Linear Representation Hypothesis with Formal Concept Analysis (FCA), showing that linear attribute directions with separating thresholds induce a concept lattice via half-space intersections. This geometry enables symbolic reasoning through geometric meet (intersection) and join (union) operations, and admits a canonical form when attribute directions are linearly independent. Experiments on WordNet sub-hierarchies provide empirical evidence that LLM embeddings encode concept lattices and their logical structure, revealing a principled bridge between continuous geometry and symbolic abstraction.
format Preprint
id arxiv_https___arxiv_org_abs_2603_01227
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle The Lattice Representation Hypothesis of Large Language Models
Xiong, Bo
Artificial Intelligence
We propose the Lattice Representation Hypothesis of large language models: a symbolic backbone that grounds conceptual hierarchies and logical operations in embedding geometry. Our framework unifies the Linear Representation Hypothesis with Formal Concept Analysis (FCA), showing that linear attribute directions with separating thresholds induce a concept lattice via half-space intersections. This geometry enables symbolic reasoning through geometric meet (intersection) and join (union) operations, and admits a canonical form when attribute directions are linearly independent. Experiments on WordNet sub-hierarchies provide empirical evidence that LLM embeddings encode concept lattices and their logical structure, revealing a principled bridge between continuous geometry and symbolic abstraction.
title The Lattice Representation Hypothesis of Large Language Models
topic Artificial Intelligence
url https://arxiv.org/abs/2603.01227