# Benchmarking Cross-Lingual Semantic Alignment in Multilingual Embeddings

arXiv: https://arxiv.org/abs/2601.09732 · HTML: https://mlsanity.com/paper/2601.09732/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 29 references that are in the map.
- First posted on arXiv 29 Dec 2025.

## Builds on (most cited first, 12 of 12)

- [UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction](https://mlsanity.com/paper/1802.03426/) (2018)
- [Text Embeddings by Weakly-Supervised Contrastive Pre-training](https://mlsanity.com/paper/2212.03533/) (2022)
- [XTREME: A Massively Multilingual Multi-task Benchmark for Evaluating Cross-lingual Generalization](https://mlsanity.com/paper/2003.11080/) (2020)
- [MTEB: Massive Text Embedding Benchmark](https://mlsanity.com/paper/2210.07316/) (2022)
- [Language-agnostic BERT Sentence Embedding](https://mlsanity.com/paper/2007.01852/) (2020)
- [Making Monolingual Sentence Embeddings Multilingual using Knowledge Distillation](https://mlsanity.com/paper/2004.09813/) (2020)
- [On the Cross-lingual Transferability of Monolingual Representations](https://mlsanity.com/paper/1910.11856/) (2019)
- [Universal Sentence Encoder](https://mlsanity.com/paper/1803.11175/) (2018)
- [Text and Code Embeddings by Contrastive Pre-Training](https://mlsanity.com/paper/2201.10005/) (2022)
- [Multilingual Universal Sentence Encoder for Semantic Retrieval](https://mlsanity.com/paper/1907.04307/) (2019)
- [Effective Parallel Corpus Mining using Bilingual Sentence Embeddings](https://mlsanity.com/paper/1807.11906/) (2018)
- [Are All Good Word Vector Spaces Isomorphic?](https://mlsanity.com/paper/2004.04070/) (2020)

## Abstract (arXiv)

With hundreds of multilingual embedding models available, practitioners lack clear guidance on which provide genuine cross-lingual semantic alignment versus task performance through language-specific patterns. Task-driven benchmarks (MTEB) may mask fundamental alignment shortcomings. We introduce Semantic Affinity (SA), a bounded (between 0 and 1) metric measuring inter-lingual to intra-lingual spread ratio using cosine distance, combined with PHATE visualization in our Semanscope framework. Benchmarking 13 models across 4 datasets (52 experiments) reveals a three-tier structure: (1) Top BERT models (LaBSE SA = 0.70, USE SA = 0.68, S-BERT SA = 0.68) achieve strong alignment via translation-pair supervision; (2) LLM embeddings plateau at SA between 0.55 and 0.61 regardless of 0.6 B to 8 B scale; (3) MLM-only BERT models (mBERT, XLM-R, SA < 0.50) fail despite more than 100 language training. Training objective, not architecture or scale, determines alignment. Oracle Bone primitives (1200 BCE) expose semantic drift-models learn corpus patterns rather than cognitive primitives. This work provides semantic benchmarking to help practitioners select quality multilingual embeddings from hundreds of available models, showing cross-lingual alignment requires explicit translation supervision, not merely model scale or multilingual data.