When Language Shapes Thought: Cross-Lingual Transfer of Factual Knowledge in Question Answering

TL;DR

This paper investigates how multilingual language models transfer factual knowledge across languages and introduces a novel prompting method, Language-to-Thought (L2T), that improves cross-lingual reasoning by aligning the model's internal thinking with the source language.

Contribution

The paper proposes the Language-to-Thought prompting method, which enhances cross-lingual factual knowledge transfer in multilingual models by aligning internal reasoning with the input language.

Findings

L2T outperforms English prompting across multiple languages and models.

Aligning internal reasoning with source language improves factual knowledge transfer.

English prompts do not always provide the best reasoning context in multilingual models.

Abstract

Multilingual large language models (LLMs) offer promising opportunities for cross-lingual information access, yet their use of factual knowledge remains highly sensitive to the input language. Prior work has addressed this through English prompting and evaluation, assuming that English-based reasoning is universally beneficial. In this work, we challenge that assumption by exploring factual knowledge transfer from non-English to English through the lens of Language and Thought Theory. We introduce Language-to-Thought (L2T) prompting, which aligns the model's internal ''thinking'' language with the source of knowledge. Across three languages and four models, L2T consistently outperforms English-based reasoning, reversing the expected advantage of English prompts. Our code is available at https://github.com/GeomeunByeol/Language2Thought.