[2603.24150] A visual observation on the geometry of UMAP projections of the difference vectors of antonym and synonym word pair embeddings

[2603.24150] A visual observation on the geometry of UMAP projections of the difference vectors of antonym and synonym word pair embeddings

arXiv - Machine Learning 3 min read

About this article

Abstract page for arXiv paper 2603.24150: A visual observation on the geometry of UMAP projections of the difference vectors of antonym and synonym word pair embeddings

Computer Science > Computation and Language arXiv:2603.24150 (cs) [Submitted on 25 Mar 2026] Title:A visual observation on the geometry of UMAP projections of the difference vectors of antonym and synonym word pair embeddings Authors:Rami Luisto View a PDF of the paper titled A visual observation on the geometry of UMAP projections of the difference vectors of antonym and synonym word pair embeddings, by Rami Luisto View PDF HTML (experimental) Abstract:Antonyms, or opposites, are sometimes defined as \emph{word pairs that have all of the same contextually relevant properties but one}. Seeing how transformer models seem to encode concepts as directions, this begs the question if one can detect ``antonymity'' in the geometry of the embedding vectors of word pairs, especially based on their difference vectors. Such geometrical studies are then naturally contrasted by comparing antonymic pairs to their opposites; synonyms. This paper started as an exploratory project on the complexity of the systems needed to detect the geometry of the embedding vectors of antonymic word pairs. What we now report is a curious ``swirl'' that appears across embedding models in a somewhat specific projection configuration. Comments: Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG) MSC classes: 68T50 (Primary) 62H30, 68T09 (Secondary) Cite as: arXiv:2603.24150 [cs.CL]   (or arXiv:2603.24150v1 [cs.CL] for this version)   https://doi.org/10.48550/arXiv.2603.24150 Focus to learn ...

Originally published on March 26, 2026. Curated by AI News.

Related Articles

UMKC Announces New Master of Science in Artificial Intelligence
Ai Infrastructure

UMKC Announces New Master of Science in Artificial Intelligence

UMKC announces a new Master of Science in Artificial Intelligence program aimed at addressing workforce demand for AI expertise, set to l...

AI News - General · 4 min ·
Llms

built an open source tool that auto generates AI context files for any codebase, 150 stars in

one of the most tedious parts of working with AI coding tools is having to manually write context files every single time. CLAUDE.md, .cu...

Reddit - Artificial Intelligence · 1 min ·
Machine Learning

[R] First open-source implementation of Hebbian fast-weight write-back for the BDH architecture

The BDH (Dragon Hatchling) paper (arXiv:2509.26507) describes a Hebbian synaptic plasticity mechanism where model weights update during i...

Reddit - Machine Learning · 1 min ·
Llms

[R] A language model built from the damped harmonic oscillator equation — no transformer blocks

I've been building a neural architecture where the only learnable transform is the transfer function of a damped harmonic oscillator: H(ω...

Reddit - Machine Learning · 1 min ·
More in Machine Learning: This Week Guide Trending

No comments

No comments yet. Be the first to comment!

Stay updated with AI News

Get the latest news, tools, and insights delivered to your inbox.

Daily or weekly digest • Unsubscribe anytime