mlsanityData to 7 Oct 2026

Hubness, Not Anisotropy, Drives Cross-Lingual Retrieval Asymmetry in Multilingual Embedding Models

arXiv 2605.26575 ↗First posted 26 May 2026cs.CLcited by 0

Not on the map. Why?

Abstract (arXiv)

Multilingual embedding models are deployed under the assumption that cross-lingual retrieval is symmetric: if a query in language A retrieves its translation in language B, the reverse should also hold. In practice it does not. Using a parallel corpus of 6,518 idiomatic and proverbial expressions in English, Bangla, Hindi, and Arabic, embedded by five production-grade encoders (Gemini, Mistral, OpenAI-L, OpenAI-S, Qwen), we formalise this failure as a deficit in mutual nearest-neighbour reciprocity and test a single mechanistic claim: among the geometric pathologies of multilingual spaces, hubness, not anisotropy, centroid drift, or magnitude, is the dominant causal driver. Across five pre-registered experiments with falsification conditions specified in advance, hub mass dominates a joint regression on reciprocity (49.5% dominance share, 1.68x the next predictor; partial R^2 = 0.302 versus 0.003 for anisotropy), while a hub-aware score correction (CSLS) closes 63.5% of the worst-to-best reciprocity gap and yields a mean within-model effect size 130x larger than surgical hub-vector ablation. The latter contrast pinpoints the mechanism: hubness is a pathology of the similarity metric, not of individual hub vectors. We resolve the well-known anisotropy-hubness paradox by showing the two are statistically dissociable, and we recommend replacing cosine similarity with CSLS as the default retrieval metric for multilingual embedding pipelines.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 17 of its 29 references that are in the map. First posted on arXiv 26 May 2026.

Builds on

17 of 29 references.

References · 17

  1. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
  2. 2019Unsupervised Cross-lingual Representation Learning at Scalecited by 3,907
  3. 2021SimCSE: Simple Contrastive Learning of Sentence Embeddingscited by 2,168
  4. 2019How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddingscited by 770
  5. 2019How multilingual is Multilingual BERT?cited by 642
  6. 2020Language-agnostic BERT Sentence Embeddingcited by 621
  7. 2020Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillationcited by 591
  8. 2017Word Translation Without Parallel Datacited by 480
  9. 2018Massively Multilingual Sentence Embeddings for Zero-Shot Cross-Lingual Transfer and Beyondcited by 434
  10. 2018SentEval: An Evaluation Toolkit for Universal Sentence Representationscited by 294
  11. 2017All-but-the-Top: Simple and Effective Postprocessing for Word Representationscited by 205
  12. 2018A robust self-learning method for fully unsupervised cross-lingual mappings of word embeddingscited by 145
5 more
  1. 2017Offline bilingual word vectors, orthogonal transformations and the inverted softmaxcited by 94
  2. 2018Loss in Translation: Learning Bilingual Word Mapping with a Retrieval Criterioncited by 69
  3. 2014Improving zero-shot learning by mitigating the hubness problemcited by 53
  4. 2019How to (Properly) Evaluate Cross-Lingual Word Embeddings: On Strong Baselines, Comparative Analyses, and Some Misconceptionscited by 48
  5. 2018Examining the Tip of the Iceberg: A Data Set for Idiom Translationcited by 34

Built on by

Not cited yet.