The Multilingual Curse at the Retrieval Layer: Evidence from Amharic

• URL: http://arxiv.org/abs/2605.24556v1
• Date: Sat, 23 May 2026 12:44:30 GMT
• Title: The Multilingual Curse at the Retrieval Layer: Evidence from Amharic
• Abstract summary: Strong zero-shot scores on multilingual benchmarks are often taken as evidence that current encoders transfer reliably across many languages.<n>We argue that this assumption breaks down for underrepresented, morphologically rich languages.<n>We compare zero-shot multilingual retrievers, Amharic-fine-tuned multilingual retrievers, and monolingual Amharic retrievers.
• Score: 49.25758237235551
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multilingual retrieval increasingly underpins cross-lingual question answering and retrieval-augmented generation. Strong zero-shot scores on multilingual benchmarks are often taken as evidence that current encoders transfer reliably across many languages. We argue that this assumption breaks down for underrepresented, morphologically rich languages, and use Amharic as a diagnostic case. Under a shared passage retrieval protocol covering dense, late-interaction, learned sparse, and cross-encoder paradigms, we compare zero-shot multilingual retrievers, Amharic-fine-tuned multilingual retrievers, and monolingual Amharic retrievers. The strongest zero-shot multilingual retriever underperforms the strongest monolingual Amharic first-stage retriever by 23% relative MRR@10. Fine-tuning two recent multilingual embedding models on the same Amharic supervision yields 32-60% relative MRR@10 gains over zero-shot, but the best Amharic-fine-tuned multilingual model remains below the strongest monolingual Amharic retriever. These findings indicate that zero-shot multilingual retrieval is not a sufficient proxy for equitable information access in the LLM era: for underrepresented languages, retrieval must be evaluated and adapted in-language rather than inferred from aggregate multilingual benchmarks. To foster future research, we publicly release the dataset, codebase, and trained models at https://github.com/rasyosef/amharic-neural-ir.

Related papers

• One Instruction Does Not Fit All: How Well Do Embeddings Align Personas and Instructions in Low-Resource Indian Languages? [1.071318785217926]
  We present a benchmark spanning 12 Indian languages and four evaluation tasks.<n>E5-Large-Instruct achieves the highest Recall@1 of 27.4% on monolingual retrieval and 20.7% on cross-lingual transfer.<n>For classification, LaBSE attains 75.3% AUROC with strong calibration.
  arXiv  Detail & Related papers  (2026-01-15T09:10:14Z)
• Investigating Language and Retrieval Bias in Multilingual Previously Fact-Checked Claim Detection [4.6738956348193]
  Large Language Models (LLMs) offer powerful capabilities for cross-lingual fact-checking.<n>LLMs often exhibit language bias, performing disproportionately better on high-resource languages such as English.<n>We present and inspect a novel concept - retrieval bias, when information retrieval systems tend to favor certain information over others.
  arXiv  Detail & Related papers  (2025-09-29T17:50:32Z)
• Synergistic Approach for Simultaneous Optimization of Monolingual, Cross-lingual, and Multilingual Information Retrieval [5.446052898856584]
  This paper proposes a novel hybrid batch training strategy to improve zero-shot retrieval performance across monolingual, cross-lingual, and multilingual settings. The approach fine-tunes multilingual language models using a mix of monolingual and cross-lingual question-answer pair batches sampled based on dataset size.
  arXiv  Detail & Related papers  (2024-08-20T04:30:26Z)
• Crosslingual Capabilities and Knowledge Barriers in Multilingual Large Language Models [62.91524967852552]
  Large language models (LLMs) are typically multilingual due to pretraining on diverse multilingual corpora.<n>But can these models relate corresponding concepts across languages, i.e., be crosslingual?<n>This study evaluates state-of-the-art LLMs on inherently crosslingual tasks.
  arXiv  Detail & Related papers  (2024-06-23T15:15:17Z)
• XAMPLER: Learning to Retrieve Cross-Lingual In-Context Examples [64.79218405438871]
  We introduce XAMPLER: Cross-Lingual Example Retrieval, a method tailored to tackle the challenge of cross-lingual in-context learning.<n>XAMPLER first trains a retriever based on Glot500, a multilingual small language model.<n>It can directly retrieve English examples as few-shot examples for in-context learning of target languages.
  arXiv  Detail & Related papers  (2024-05-08T15:13:33Z)
• Unsupervised Multilingual Dense Retrieval via Generative Pseudo Labeling [32.10366004426449]
  This paper introduces UMR, an Unsupervised dense Multilingual Retriever trained without any paired data. We propose a two-stage framework which iteratively improves the performance of multilingual dense retrievers.
  arXiv  Detail & Related papers  (2024-03-06T07:49:06Z)
• Towards Better Monolingual Japanese Retrievers with Multi-Vector Models [0.0]
  In Japanese, the best performing deep-learning based retrieval approaches rely on multilingual dense embedders. We introduce JaColBERT, a family of multi-vector retrievers trained on two magnitudes fewer data than their multilingual counterparts.
  arXiv  Detail & Related papers  (2023-12-26T18:07:05Z)
• Leveraging LLMs for Synthesizing Training Data Across Many Languages in Multilingual Dense Retrieval [56.65147231836708]
  We develop SWIM-IR, a synthetic retrieval training dataset containing 33 languages for fine-tuning multilingual dense retrievers. SAP assists the large language model (LLM) in generating informative queries in the target language. Our models, called SWIM-X, are competitive with human-supervised dense retrieval models.
  arXiv  Detail & Related papers  (2023-11-10T00:17:10Z)
• The Belebele Benchmark: a Parallel Reading Comprehension Dataset in 122 Language Variants [80.4837840962273]
  We present Belebele, a dataset spanning 122 language variants. This dataset enables the evaluation of text models in high-, medium-, and low-resource languages.
  arXiv  Detail & Related papers  (2023-08-31T17:43:08Z)
• XCOPA: A Multilingual Dataset for Causal Commonsense Reasoning [68.57658225995966]
  Cross-lingual Choice of Plausible Alternatives (XCOPA) is a typologically diverse multilingual dataset for causal commonsense reasoning in 11 languages. We evaluate a range of state-of-the-art models on this novel dataset, revealing that the performance of current methods falls short compared to translation-based transfer.
  arXiv  Detail & Related papers  (2020-05-01T12:22:33Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.