Paper page - Code-Switching Information Retrieval: Benchmarks, Analysis, and the Limits of Current Retrievers

Source: https://huggingface.co/papers/2604.17632

Abstract

Code-switching poses significant challenges for information retrieval systems, revealing performance bottlenecks and embedding space divergences that current multilingual approaches cannot fully address.

Code-switchingis a pervasive linguistic phenomenon in global communication, yet moderninformation retrievalsystems remain predominantly designed for, and evaluated within, monolingual contexts. To bridge this critical disconnect, we present a holistic study dedicated tocode-switchingIR. We introduceCSR-L(Code-SwitchingRetrieval benchmark-Lite), constructing a dataset via human annotation to capture the authentic naturalness of mixed-language queries. Our evaluation across statistical, dense, and late-interaction paradigms reveals thatcode-switchingacts as a fundamental performance bottleneck, degrading the effectiveness of even robustmultilingual models. We demonstrate that this failure stems from substantial divergence in theembedding spacebetween pure and code-switched text. Scaling this investigation, we proposeCS-MTEB, a comprehensive benchmark covering 11 diverse tasks, where we observe performance declines of up to 27%. Finally, we show that standard multilingual techniques likevocabulary expansionare insufficient to resolve these deficits completely. These findings underscore the fragility of current systems and establishcode-switchingas a crucial frontier for future IR optimization.

View arXiv pageView PDFGitHub0Add to collection

Get this paper in your agent:

hf papers read 2604\.17632

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2604.17632 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2604.17632 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2604.17632 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.