Multilingual Prompting for Improving LLM Generation Diversity

2025/05/21 by Qihan Wang, Shidong Pan, Wang, Qihan +5 · 1 voice · 8 citations
Computer Science · Social Sciences · #Computation and Language (cs.CL) #Computers and Society (cs.CY) #Digital Rights Management and Security #FOS: Computer and information sciences #Wikis in Education and Collaboration #cs.CL #cs.CY

paper · pdf · doi:10.48550/arxiv.2505.15229

openalex publication_date 2025/05/21 · arxiv published 2025/05/21 · arxiv updated 2025/09/27 · openalex created_date 2025/10/19 · openalex updated_date 2026/07/28

Abstract

Large Language Models (LLMs) are known to lack cultural representation and overall diversity in their generations, from expressing opinions to answering factual questions. To mitigate this problem, we propose multilingual prompting: a prompting method which generates several variations of a base prompt with added cultural and linguistic cues from several cultures, generates responses, and then combines the results. Building on evidence that LLMs have language-specific knowledge, multilingual prompting seeks to increase diversity by activating a broader range of cultural knowledge embedded in model training data. Through experiments across multiple models (GPT-4o, GPT-4o-mini, LLaMA 70B, and LLaMA 8B), we show that multilingual prompting consistently outperforms existing diversity-enhancing techniques such as high-temperature sampling, step-by-step recall, and persona prompting. Further analyses show that the benefits of multilingual prompting vary between high and low resource languages and across model sizes, and that aligning the prompting language with cultural cues reduces hallucination about culturally-specific information.

Citations

• Language Models' Factuality Depends on the Language of Inquiry
• A Note on Shumailov et al. (2024): `AI Models Collapse When Trained on Recursively Generated Data'
• Evaluating Cultural Adaptability of a Large Language Model via Simulation of Synthetic Personas
• Cultural Conditioning or Placebo? On the Effectiveness of Socio-Demographic Prompting
• Large language models that replace human participants can harmfully misportray and flatten identity groups
• Improving Diversity of Demographic Representation in Large Language Models via Collective-Critiques and Self-Voting
• Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting
• Sensitivity, Performance, Robustness: Deconstructing the Effect of Sociodemographic Prompting
• Does Writing with Language Models Reduce Content Diversity?
• Marked Personas: Using Natural Language Prompts to Measure Stereotypes in Language Models
• Separating form and meaning: Using self-consistency to quantify task understanding across multiple senses
• Whose Opinions Do Language Models Reflect?
• Picking on the Same Person: Does Algorithmic Monoculture lead to Outcome Homogenization?
• Language Models are Multilingual Chain-of-Thought Reasoners
• Social Chemistry 101: Learning to Reason about Social and Moral Norms
• Evaluating the Evaluation of Diversity in Natural Language Generation
• HellaSwag: Can a Machine Really Finish Your Sentence?

Cited by

• CIEGAD: Cluster-Conditioned Interpolative and Extrapolative Framework for Geometry-Aware and Domain-Aligned Data Augmentation
• CrowdLLM: Building LLM-Based Digital Populations Augmented with Generative Models
• Addressing LLM Diversity by Infusing Random Concepts
• Optimizing Diversity and Quality through Base-Aligned Model Collaboration
• PromptGuard at BLP-2025 Task 1: A Few-Shot Classification Framework Using Majority Voting and Keyword Similarity for Bengali Hate Speech Detection
• Camellia: Benchmarking Cultural Biases in LLMs for Asian Languages
• ConlangCrafter: Constructing Languages with a Multi-Hop LLM Pipeline
• Cross-cultural evaluation of taste-sound correspondences in AI-generated music

Discussions

• Preprint: Multilingual Prompting for Improving #LLM Generation #Diversity arxiv.org/abs/2505.15229 #ai [bsky, 0 points, 0 comments]

Related

• Not All Languages Are Created Equal in LLMs: Improving Multilingual Capability by Cross-Lingual-Thought Prompting
• Teaching LLMs to Abstain across Languages via Multilingual Feedback
• Disentangling Language and Culture for Evaluating Multilingual Large Language Models
• CulFiT: A Fine-grained Cultural-aware LLM Training Paradigm via Multilingual Critique Data Synthesis
• CulturalTeaming: AI-Assisted Interactive Red-Teaming for Challenging LLMs' (Lack of) Multicultural Knowledge