The Medium Is Not the Message: Deconfounding Document Embeddings via Linear Concept Erasure

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Fan, Yu, Tian, Yang, Ravfogel, Shauli, Sachan, Mrinmaya, Ash, Elliott, Hoyle, Alexander
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909803741184000
author Fan, Yu
Tian, Yang
Ravfogel, Shauli
Sachan, Mrinmaya
Ash, Elliott
Hoyle, Alexander
author_facet Fan, Yu
Tian, Yang
Ravfogel, Shauli
Sachan, Mrinmaya
Ash, Elliott
Hoyle, Alexander
contents Embedding-based similarity metrics between text sequences can be influenced not just by the content dimensions we most care about, but can also be biased by spurious attributes like the text's source or language. These document confounders cause problems for many applications, but especially those that need to pool texts from different corpora. This paper shows that a debiasing algorithm that removes information about observed confounders from the encoder representations substantially reduces these biases at a minimal computational cost. Document similarity and clustering metrics improve across every embedding variant and task we evaluate -- often dramatically. Interestingly, performance on out-of-distribution benchmarks is not impacted, indicating that the embeddings are not otherwise degraded.
format Preprint
id arxiv_https___arxiv_org_abs_2507_01234
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle The Medium Is Not the Message: Deconfounding Document Embeddings via Linear Concept Erasure
Fan, Yu
Tian, Yang
Ravfogel, Shauli
Sachan, Mrinmaya
Ash, Elliott
Hoyle, Alexander
Computation and Language
Embedding-based similarity metrics between text sequences can be influenced not just by the content dimensions we most care about, but can also be biased by spurious attributes like the text's source or language. These document confounders cause problems for many applications, but especially those that need to pool texts from different corpora. This paper shows that a debiasing algorithm that removes information about observed confounders from the encoder representations substantially reduces these biases at a minimal computational cost. Document similarity and clustering metrics improve across every embedding variant and task we evaluate -- often dramatically. Interestingly, performance on out-of-distribution benchmarks is not impacted, indicating that the embeddings are not otherwise degraded.
title The Medium Is Not the Message: Deconfounding Document Embeddings via Linear Concept Erasure
topic Computation and Language
url https://arxiv.org/abs/2507.01234