LayerFlow: Layer-wise Exploration of LLM Embeddings using Uncertainty-aware Interlinked Projections

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sevastjanova, Rita, Gerling, Robin, Spinner, Thilo, El-Assady, Mennatallah
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912328576925696
author Sevastjanova, Rita
Gerling, Robin
Spinner, Thilo
El-Assady, Mennatallah
author_facet Sevastjanova, Rita
Gerling, Robin
Spinner, Thilo
El-Assady, Mennatallah
contents Large language models (LLMs) represent words through contextual word embeddings encoding different language properties like semantics and syntax. Understanding these properties is crucial, especially for researchers investigating language model capabilities, employing embeddings for tasks related to text similarity, or evaluating the reasons behind token importance as measured through attribution methods. Applications for embedding exploration frequently involve dimensionality reduction techniques, which reduce high-dimensional vectors to two dimensions used as coordinates in a scatterplot. This data transformation step introduces uncertainty that can be propagated to the visual representation and influence users' interpretation of the data. To communicate such uncertainties, we present LayerFlow - a visual analytics workspace that displays embeddings in an interlinked projection design and communicates the transformation, representation, and interpretation uncertainty. In particular, to hint at potential data distortions and uncertainties, the workspace includes several visual components, such as convex hulls showing 2D and HD clusters, data point pairwise distances, cluster summaries, and projection quality metrics. We show the usability of the presented workspace through replication and expert case studies that highlight the need to communicate uncertainty through multiple visual components and different data perspectives.
format Preprint
id arxiv_https___arxiv_org_abs_2504_10504
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LayerFlow: Layer-wise Exploration of LLM Embeddings using Uncertainty-aware Interlinked Projections
Sevastjanova, Rita
Gerling, Robin
Spinner, Thilo
El-Assady, Mennatallah
Computation and Language
Graphics
Large language models (LLMs) represent words through contextual word embeddings encoding different language properties like semantics and syntax. Understanding these properties is crucial, especially for researchers investigating language model capabilities, employing embeddings for tasks related to text similarity, or evaluating the reasons behind token importance as measured through attribution methods. Applications for embedding exploration frequently involve dimensionality reduction techniques, which reduce high-dimensional vectors to two dimensions used as coordinates in a scatterplot. This data transformation step introduces uncertainty that can be propagated to the visual representation and influence users' interpretation of the data. To communicate such uncertainties, we present LayerFlow - a visual analytics workspace that displays embeddings in an interlinked projection design and communicates the transformation, representation, and interpretation uncertainty. In particular, to hint at potential data distortions and uncertainties, the workspace includes several visual components, such as convex hulls showing 2D and HD clusters, data point pairwise distances, cluster summaries, and projection quality metrics. We show the usability of the presented workspace through replication and expert case studies that highlight the need to communicate uncertainty through multiple visual components and different data perspectives.
title LayerFlow: Layer-wise Exploration of LLM Embeddings using Uncertainty-aware Interlinked Projections
topic Computation and Language
Graphics
url https://arxiv.org/abs/2504.10504