When English Isn't the Best Teacher: Source Language Effects in Cross-Lingual In-Context Learning

2026-06-16 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Expert, quick

Summary

A broad empirical study investigates cross-lingual transfer in In-Context Learning (ICL), challenging the common assumption that insights from supervised fine-tuning contexts directly apply. The research rigorously evaluates how to choose optimal source languages for cross-lingual ICL, covering seven distinct tasks, six different models, and a typologically diverse set of languages. Additionally, the study analyzes language confusion, identified as a key obstacle for generative tasks within cross-lingual ICL. Its findings demonstrate that conventional expectations, largely based on fine-tuning, do not consistently hold true in the ICL regime, pointing instead to alternative heuristics for more effective source language selection.

Key takeaway

For NLP engineers developing cross-lingual applications with In-Context Learning, you should re-evaluate your assumptions about source language selection. The study indicates that strategies effective in fine-tuning may not yield optimal results in ICL. Focus on exploring alternative heuristics for source language choice and specifically address potential language confusion when working with generative cross-lingual ICL tasks to improve performance and reliability.

Key insights

ICL cross-lingual transfer differs from fine-tuning, requiring new source language selection heuristics.

Principles

• Fine-tuning insights don't always apply to ICL.
• Source language choice impacts cross-lingual ICL.
• Language confusion is an ICL generative task obstacle.

Method

Conducted a broad empirical study across seven tasks, six models, and diverse languages, analyzing language confusion.

In practice

• Re-evaluate source language selection for ICL.
• Consider language confusion in generative ICL.

Topics

• Cross-lingual Transfer
• In-Context Learning
• Source Language Selection
• Multilingual NLP
• Language Confusion
• Generative AI
• Fine-tuning

Best for: AI Scientist, NLP Engineer, Research Scientist

Related on AIssential

• Revisiting the Systematicity in Negation in the Era of In-Context Learning — LLMs partially grasp negation via in-context learning, but struggle with robust scope recognition and function vector composition for it.
• Source-Grounded Semantic Reinforcement Learning for Low-Resource Target-Language Generation — SG-SRL uses source-language monolingual data with cross-lingual semantic RL to improve low-resource target-language generation, recovering fluency with a small parallel corpus.
• KIT's Submission to Cross-Lingual Voice Cloning in IWSLT 2026 — Explicit language tags and lexical matching significantly improve cross-lingual voice cloning performance and reduce accent leakage.
• Event and Entity Coreference Across Five Languages: Effects of Context and Referring Expression — Crosslingual coreference biases show personal pronouns favor entities, demonstratives favor events, influenced by antecedent complexity.
• Lost in the Tower of Babel: The Adverse Effects of Incidental Multilingualism in LLMs — Incidental multilingualism in LLMs leads to brittle, unequal, and opaque cross-lingual performance.
• Multilingual Idioms in Sentences and Conversations Across High-, Medium-, and Low-Resource Languages — Idiomatic expressions challenge multilingual NLP, with literal meanings and low-resource languages posing significant hurdles.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.