Location Not Found: Exposing Implicit Local and Global Biases in Multilingual LLMs

TL;DR

This paper introduces LocQA, a multilingual test set to evaluate biases in LLMs, revealing a US-centric global bias and demographic preferences within languages, especially after instruction tuning.

Contribution

The work presents a novel benchmark, LocQA, to quantify implicit local and global biases in multilingual LLMs, highlighting biases amplified by instruction tuning.

Findings

Models show a US-centric bias in answers across languages.

Instruction tuning increases the global US bias.

Models prioritize larger population locales within the same language.

Abstract

Multilingual large language models (LLMs) have minimized the fluency gap between languages. This advancement, however, exposes models to the risk of biased behavior, as knowledge and norms may propagate across languages. In this work, we aim to quantify models' inter- and intra-lingual biases, via their ability to answer locale-ambiguous questions. To this end, we present LocQA, a test set containing 2,156 questions in 12 languages, referring to various locale-dependent facts such as laws, dates, and measurements. The questions do not contain indications of the locales they relate to, other than the querying language itself. LLMs' responses to LocQA locale-ambiguous questions thus reveal models' implicit priors. We used LocQA to evaluate 32 models, and detected two types of structural biases. Inter-lingually, we show a global bias towards answers relevant to the US-locale, even when models are asked in languages other than English. Moreover, we discovered that this global bias is exacerbated in models that underwent instruction tuning, compared to their base counterparts. Intra-lingually, we show that when multiple locales are relevant for the same language, models act as demographic probability engines, prioritizing locales with larger populations. Taken together, insights from LocQA may help in shaping LLMs' desired local behavior, and in quantifying the impact of various training phases on different kinds of biases.