Lost in Translation? A Comparative Study on the Cross-Lingual Transfer of Composite Harms

• URL: http://arxiv.org/abs/2602.07963v1
• Date: Sun, 08 Feb 2026 13:22:50 GMT
• Title: Lost in Translation? A Comparative Study on the Cross-Lingual Transfer of Composite Harms
• Abstract summary: Most safety evaluations of large language models (LLMs) remain anchored in English.<n>Some types of harm survive translation almost intact, while others distort or disappear.<n>We introduce CompositeHarm, a translation-based benchmark to examine how safety alignment holds up as both syntax and semantics shift.
• Score: 0.5376203747548287
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Most safety evaluations of large language models (LLMs) remain anchored in English. Translation is often used as a shortcut to probe multilingual behavior, but it rarely captures the full picture, especially when harmful intent or structure morphs across languages. Some types of harm survive translation almost intact, while others distort or disappear. To study this effect, we introduce CompositeHarm, a translation-based benchmark designed to examine how safety alignment holds up as both syntax and semantics shift. It combines two complementary English datasets, AttaQ, which targets structured adversarial attacks, and MMSafetyBench, which covers contextual, real-world harms, and extends them into six languages: English, Hindi, Assamese, Marathi, Kannada, and Gujarati. Using three large models, we find that attack success rates rise sharply in Indic languages, especially under adversarial syntax, while contextual harms transfer more moderately. To ensure scalability and energy efficiency, our study adopts lightweight inference strategies inspired by edge-AI design principles, reducing redundant evaluation passes while preserving cross-lingual fidelity. This design makes large-scale multilingual safety testing both computationally feasible and environmentally conscious. Overall, our results show that translated benchmarks are a necessary first step, but not a sufficient one, toward building grounded, resource-aware, language-adaptive safety systems.

Related papers

• IndicSafeEval: Safety Robustness of Large Language Models under Multilingual Persuasive Jailbreak Attacks [33.08360320836576]
  We introduce IndicSafeEval, a persuasion-based jailbreak evaluation framework for Indian languages.<n>Our benchmark combines ten safety critical content categories with six human-like persuasive strategies across four different Indian languages.<n>Our analysis shows that the model does not behave equally safely across all languages and prompt styles.
  arXiv  Detail & Related papers  (2026-09-03T12:54:44Z)
• The Illusion of Cross-Lingual Safety in Low-Resource Languages [7.471112976694495]
  Safety alignment in large language models (LLMs) is largely developed in English, assuming these safeguards generalize across multilingual settings.<n>We investigate cross-lingual safety transfer in four African languages, Twi, Hausa, Amharic, and Swahili, using LoDNA, a new safety dataset that pairs literal translations with culturally localized prompts.<n>Our experimental results show that cross-lingual safety transfer is severely limited; harmful prompts retain less than 10% of the English refusal signal across most language-model pairs.
  arXiv  Detail & Related papers  (2026-08-11T17:05:26Z)
• The Heterogeneous Safety Impacts of Benign Multilingual Fine-Tuning [13.578131927232292]
  Fine-tuning a large language model can increase its tendency to respond to unsafe adversarial prompts.<n>We present the first comprehensive empirical study of this phenomenon in multilingual settings by fine-tuning Llama-3.2, Qwen3, and Gemma-3 models.<n>We find that safety outcomes are highly sensitive to both the choice of fine-tuning language and the evaluation language.
  arXiv  Detail & Related papers  (2026-06-27T10:05:36Z)
• Sycophancy as a Multilingual Alignment Failure: How Safety Degrades Across Languages, Topics, and Models [3.2555924024638667]
  Safety-aligned large language models often exhibit sycophancy, which is the tendency to affirm users' opinions regardless of factual accuracy.<n>We present the first large-scale, multi-model evaluation of cross-lingual sycophancy, benchmarking textbfsix instruction-tuned models across textbf38 languages and textbf33 topic categories.
  arXiv  Detail & Related papers  (2026-06-07T04:50:40Z)
• Exploring Adversarial Robustness and Safety Alignment in Multilingual Multi-Modal Large Language Models [37.24612885977265]
  Multimodal Large Language Models integrate visual perception into language reasoning, introducing a continuous attack surface susceptible to adversarial attacks.<n>Prior work on MLLM has focused largely on English-centric tasks, leaving multilingual behaviour unexplored.<n>We study adversarial robustness and multimodal safety across 12 diverse languages, evaluating open-source MLLMs that acquire multilingual capability through instruction tuning.
  arXiv  Detail & Related papers  (2026-06-02T15:42:10Z)
• Multilingual Refusal Alignment for Safer Large Language Models [53.64286756804503]
  We investigate whether single-language alignment transfers cross-lingually, how language consistency is preserved during training, and the resulting trade-offs with general knowledge capabilities.<n>We introduce RefusEU, a novel refusal alignment dataset covering 12 European languages, including a dedicated test set for evaluating current state-of-the-art models.<n>Our controlled Direct Preference Optimization (DPO) experiments provide two key insights: aligning models exclusively in English is insufficient to ensure cross-lingual safety, even for the same harm categories, whereas training on multilingual datasets can improve safety without degrading general performance, as measured by the Global MMLU benchmark.
  arXiv  Detail & Related papers  (2026-04-24T09:29:14Z)
• LASA: Language-Agnostic Semantic Alignment at the Semantic Bottleneck for LLM Safety [88.98698230989186]
  Large language models (LLMs) often demonstrate strong safety performance in high-resource languages, yet exhibit severe vulnerabilities in low-resource languages.<n>We attribute this gap to a mismatch between language-agnostic semantic understanding ability and language-dominant safety alignment biased toward high-resource languages.<n>We propose Language-Agnostic Semantic Alignment (LASA), which anchors safety alignment directly in semantic bottlenecks.
  arXiv  Detail & Related papers  (2026-04-13T15:59:50Z)
• When Meanings Meet: Investigating the Emergence and Quality of Shared Concept Spaces during Multilingual Language Model Training [57.230355403478995]
  We investigate the development of language-agnostic concept spaces during pretraining of EuroLLM.<n>We find that shared concept spaces emerge early and continue to refine, but that alignment with them is language-dependent.<n>In contrast to prior work, our fine-grained manual analysis reveals that some apparent gains in translation quality reflect shifts in behavior.
  arXiv  Detail & Related papers  (2026-01-30T11:23:01Z)
• Multilinguality as Sense Adaptation [24.548610248136352]
  SENse-based Symmetric Interlingual Alignment (SENSIA)<n>We introduce SENse-based Symmetric Interlingual Alignment (SENSIA)<n>It adapts a Backpack language model from one language to another by explicitly aligning sense-level mixtures and contextual representations on parallel data.
  arXiv  Detail & Related papers  (2026-01-15T11:44:01Z)
• On the Entity-Level Alignment in Crosslingual Consistency [62.33186691736433]
  SubSub and SubInj integrate English translations of subjects into prompts across languages, leading to substantial gains in factual recall accuracy and consistency.<n>These interventions reinforce the entity representation alignment in the conceptual space through model's internal pivot-language processing.
  arXiv  Detail & Related papers  (2025-10-11T16:26:50Z)
• Toxicity-Aware Few-Shot Prompting for Low-Resource Singlish Translation [3.7678366606419345]
  Translating toxic content between low-resource language pairs poses challenges due to scarce parallel data and safety filters that sanitize offensive expressions.<n>We propose a two-stage framework for toxicity-preserving translation, demonstrated on a code-mixed Singlish safety corpus.<n>By positioning Singlish as a testbed for inclusive NLP, we underscore the importance of preserving sociolinguistic nuance in real-world applications.
  arXiv  Detail & Related papers  (2025-07-16T06:58:02Z)
• MrGuard: A Multilingual Reasoning Guardrail for Universal LLM Safety [56.77103365251923]
  Large Language Models (LLMs) are susceptible to adversarial attacks such as jailbreaking.<n>This vulnerability is exacerbated in multilingual settings, where multilingual safety-aligned data is often limited.<n>We introduce a multilingual guardrail with reasoning for prompt classification.
  arXiv  Detail & Related papers  (2025-04-21T17:15:06Z)
• Translation Errors Significantly Impact Low-Resource Languages in Cross-Lingual Learning [26.49647954587193]
  In this work, we find that translation inconsistencies do exist and they disproportionally impact low-resource languages in XNLI. To identify such inconsistencies, we propose measuring the gap in performance between zero-shot evaluations on the human-translated and machine-translated target text. We also corroborate that translation errors exist for two target languages, namely Hindi and Urdu, by doing a manual reannotation of human-translated test instances.
  arXiv  Detail & Related papers  (2024-02-03T08:22:51Z)
• Lost In Translation: Generating Adversarial Examples Robust to Round-Trip Translation [66.33340583035374]
  We present a comprehensive study on the robustness of current text adversarial attacks to round-trip translation. We demonstrate that 6 state-of-the-art text-based adversarial attacks do not maintain their efficacy after round-trip translation. We introduce an intervention-based solution to this problem, by integrating Machine Translation into the process of adversarial example generation.
  arXiv  Detail & Related papers  (2023-07-24T04:29:43Z)
• Adversarial GLUE: A Multi-Task Benchmark for Robustness Evaluation of Language Models [86.02610674750345]
  Adversarial GLUE (AdvGLUE) is a new multi-task benchmark to explore and evaluate the vulnerabilities of modern large-scale language models under various types of adversarial attacks. We apply 14 adversarial attack methods to GLUE tasks to construct AdvGLUE, which is further validated by humans for reliable annotations. All the language models and robust training methods we tested perform poorly on AdvGLUE, with scores lagging far behind the benign accuracy.
  arXiv  Detail & Related papers  (2021-11-04T12:59:55Z)
• AM2iCo: Evaluating Word Meaning in Context across Low-ResourceLanguages with Adversarial Examples [51.048234591165155]
  We present AM2iCo, Adversarial and Multilingual Meaning in Context. It aims to faithfully assess the ability of state-of-the-art (SotA) representation models to understand the identity of word meaning in cross-lingual contexts. Results reveal that current SotA pretrained encoders substantially lag behind human performance.
  arXiv  Detail & Related papers  (2021-04-17T20:23:45Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.