Lune

EMNLP2024Top-tier venue

The Multilingual Alignment Prism: Aligning Global and Local Preferences to Reduce Harm

Aakanksha, Arash Ahmadian, Beyza Ermis, Seraphina Goldfarb-Tarrant, Julia Kreutzer, Marzieh Fadaee, Sara Hooker

Abstract

A key concern with the concept of alignment is the implicit question of alignment to what? AI systems are increasingly used across the world, yet safety alignment is often focused on homogeneous monolingual settings. Additionally, preference training and safety measures often overfit to harms common in Western-centric datasets. Here, we explore the viability of different alignment approaches when balancing dual objectives: addressing and optimizing for a non-homogeneous set of languages and cultural preferences while minimizing both global and local harms. We collect the first set of human-annotated red-teaming prompts 1 in different languages distinguishing between global and local harm, which serve as a laboratory for understanding the reliability of alignment techniques when faced with preference distributions that are non-stationary across geographies and languages. While this setting is seldom covered by the literature to date, which primarily centers on English harm mitigation, it captures real-world interactions with AI systems around the world. We establish a new precedent for state-of-the-art alignment techniques across 6 languages with minimal degradation in general performance. Our work provides important insights into cross-lingual transfer and novel optimization approaches to safeguard AI systems designed to serve global populations.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

24 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Cultivating Pluralism In Algorithmic Monoculture: The Community Alignment Dataset cites Direct Preference Optimization: Your Language Model is Secretly a Reward Model
  • Kaleidoscope: In-language Exams for Massively Multilingual Vision Evaluation cites Global MMLU: Understanding and Addressing Cultural and Linguistic Biases in Multilingual Evaluation
  • MPO: Multilingual Safety Alignment via Reward Gap Optimization cites Direct Preference Optimization: Your Language Model is Secretly a Reward Model
  • MPO: Multilingual Safety Alignment via Reward Gap Optimization cites Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers24

  1. Global MMLU: Understanding and Addressing Cultural and Linguistic Biases in Multilingual EvaluationShivalika Singh, Angelika Romanou, Clémentine Fourrier, David Ifeoluwa Adelani et al.ACL 2025 · 144 citations
  2. Detecting High-Stakes Interactions with Activation ProbesAlex McKenzie, Urja Pawar, Phil Blandfort, William Bankes et al.NeurIPS 2025 · 52 citations
  3. Cultivating Pluralism In Algorithmic Monoculture: The Community Alignment DatasetLily H Zhang, Smitha Milli, Karen Long Jusko, Jonathan Smith et al.ICLR 2026 · 41 citations
  4. MPO: Multilingual Safety Alignment via Reward Gap OptimizationWeixiang Zhao, Yulin Hu, Yang Deng, Tongtong Wu et al.ACL 2025 · 12 citations
  5. Kaleidoscope: In-language Exams for Massively Multilingual Vision EvaluationIsrafel Salazar, Manuel Fernández Burda, Shayekh Bin Islam, Arshia Soltani Moakhar et al.ICLR 2026 · 8 citations

Builds on22

  1. Chain-of-Thought Prompting Elicits Reasoning in Large Language ModelsJason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma et al.NeurIPS 2022 · 22,562 citations
  2. Direct Preference Optimization: Your Language Model is Secretly a Reward ModelRafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D. Manning et al.NeurIPS 2023 · 10,924 citations
  3. WizardCoder: Empowering Code Large Language Models with Evol-InstructZiyang Luo, Can Xu, Pu Zhao, Qingfeng Sun et al.ICLR 2024 · 945 citations
  4. Are aligned neural networks adversarially aligned?Nicholas Carlini, Milad Nasr, Christopher A. Choquette-Choo, Matthew Jagielski et al.NeurIPS 2023 · 412 citations
  5. Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow InstructionsFederico Bianchi, Mirac Suzgun, Giuseppe Attanasio, Paul Röttger et al.ICLR 2024 · 373 citations

Related papers

  1. Multilingual Refusal Alignment for Safer Large Language ModelsAleksandra Krasnodebska, Wojciech Kusa, Aldo LipaniACL Findings 2026
  2. PLLuM-Align: Polish Preference Dataset for Large Language Model AlignmentKarolina Seweryn, Anna Kolos, Agnieszka Karlinska, Katarzyna Lorenc et al.EMNLP 2025
  3. Unintended Impacts of LLM Alignment on Global RepresentationMichael J. Ryan, William Barr Held, Diyi YangACL 2024
  4. The Language Barrier: Dissecting Safety Challenges of LLMs in Multilingual ContextsLingfeng Shen, Weiting Tan, Sihao Chen, Yunmo Chen et al.ACL Findings 2024 · 25 citations
  5. All Languages Matter: On the Multilingual Safety of LLMsWenxuan Wang, Zhaopeng Tu, Chang Chen, Youliang Yuan et al.ACL Findings 2024 · 21 citations
Dusk over the sea between two cliffs drawn in fine vertical lines