Evaluating and Mitigating Linguistic Discrimination in Large Language Models

TL;DR

This paper investigates linguistic discrimination in large language models, revealing disparities in safety and quality across languages, and proposes LDFighter, a method to improve consistency and mitigate biases.

Contribution

The study provides a comprehensive analysis of language-based disparities in LLM outputs and introduces LDFighter, a novel similarity-based voting method to enhance fairness and response quality.

Findings

LLMs are more aligned and safer in English, French, Russian, and Spanish.

Languages like Bengali, Georgian, Nepali, and Maithili show higher jailbreak success rates.

LDFighter reduces jailbreak success and improves response quality across languages.

Abstract

By training on text in various languages, large language models (LLMs) typically possess multilingual support and demonstrate remarkable capabilities in solving tasks described in different languages. However, LLMs can exhibit linguistic discrimination due to the uneven distribution of training data across languages. That is, LLMs are hard to keep the consistency of responses when faced with the same task but depicted in different languages. In this study, we first explore the consistency in the LLMs' outputs responding to queries in various languages from two aspects: safety and quality. We conduct this analysis with two datasets (AdvBench and NQ) based on four LLMs (Llama2-13b, Gemma-7b, GPT-3.5-turbo and Gemini-pro). The results show that LLMs exhibit stronger human alignment capabilities with queries in English, French, Russian, and Spanish (only 1.04\% of harmful queries successfully jailbreak on average) compared to queries in Bengali, Georgian, Nepali and Maithili (27.7\% of harmful queries jailbreak successfully on average). Moreover, for queries in English, Danish, Czech and Slovenian, LLMs tend to produce responses with a higher quality (with 0.1494 $F_1$ score on average) compared to the other languages. Upon these findings, we propose LDFighter, a similarity-based voting, to mitigate the linguistic discrimination in LLMs. LDFighter ensures consistent service for different language speakers. We evaluate LDFighter with both benign queries and harmful queries. The results show that LDFighter not only significantly reduces the jailbreak success rate but also improve the response quality on average, demonstrating its effectiveness.