FLiP: Towards understanding and interpreting multimodal multilingual sentence embeddings

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kesiraju, Santosh, Yusuf, Bolaji, Sedláček, Šimon, Plchot, Oldřich, Schwarz, Petr
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915945622011904
author Kesiraju, Santosh
Yusuf, Bolaji
Sedláček, Šimon
Plchot, Oldřich
Schwarz, Petr
author_facet Kesiraju, Santosh
Yusuf, Bolaji
Sedláček, Šimon
Plchot, Oldřich
Schwarz, Petr
contents This paper presents factorized linear projection (FLiP) models for understanding pretrained sentence embedding spaces. We train FLiP models to recover the lexical content from multilingual (LaBSE), multimodal (SONAR) and API-based (Gemini) sentence embedding spaces in several high- and mid-resource languages. We show that FLiP can recall more than 75% of lexical content from the embeddings, significantly outperforming existing non-factorized baselines. Using this as a diagnostic tool, we uncover the modality and language biases across the selected sentence encoders and provide practitioners with intrinsic insights about the encoders without relying on conventional downstream evaluation tasks. Our implementation is public https://github.com/BUTSpeechFIT/FLiP.
format Preprint
id arxiv_https___arxiv_org_abs_2604_18109
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle FLiP: Towards understanding and interpreting multimodal multilingual sentence embeddings
Kesiraju, Santosh
Yusuf, Bolaji
Sedláček, Šimon
Plchot, Oldřich
Schwarz, Petr
Computation and Language
Sound
This paper presents factorized linear projection (FLiP) models for understanding pretrained sentence embedding spaces. We train FLiP models to recover the lexical content from multilingual (LaBSE), multimodal (SONAR) and API-based (Gemini) sentence embedding spaces in several high- and mid-resource languages. We show that FLiP can recall more than 75% of lexical content from the embeddings, significantly outperforming existing non-factorized baselines. Using this as a diagnostic tool, we uncover the modality and language biases across the selected sentence encoders and provide practitioners with intrinsic insights about the encoders without relying on conventional downstream evaluation tasks. Our implementation is public https://github.com/BUTSpeechFIT/FLiP.
title FLiP: Towards understanding and interpreting multimodal multilingual sentence embeddings
topic Computation and Language
Sound
url https://arxiv.org/abs/2604.18109