The Shibboleth Effect: Cross-lingual behavioral skew in frontier LLMs under adversarial geopolitical simulation
Researchers introduce the 'Shibboleth Effect' — systematic behavioral differences in LLMs when operating in different languages — and audit six frontier models (GPT-4o, Llama-4, Mistral-Large, Gemini-3.1-Pro, Qwen3.6-Plus, DeepSeek-R1) using a synthetic maritime territorial dispute wargame played in English versus Turkish. Results are heterogeneous: Llama-4 becomes significantly more coercive in Turkish while Gemini-3.1-Pro and DeepSeek-R1 become less so, and GPT-4o shows no detectable shift. The study identifies two candidate buffering mechanisms — chain-of-thought institutional anchoring and multilingual RLHF alignment — with direct implications for deploying LLMs in diplomatic or crisis-management contexts.
Related guides (4)
Related events (8)
Geopolitical Bias in LLMs Originates in Post-Training, Not Pre-Training Data
A study testing seven open-weight LLM pairs (base vs. chat models) across seven labs finds that geopolitical bias is introduced during post-training rather than inherited from pre-training data. Six of seven labs showed post-training shifts favoring the developer's home country or region, with Alibaba's Qwen 2.5 showing the most extreme shift (18x increase in China-favourability log-odds). The effect is also language-dependent: Mistral becomes pro-France only under French prompting. The authors argue this implicates alignment and RLHF processes as active shapers of geopolitical perspective, calling for greater transparency and auditing of post-training pipelines.
Adversarial robustness and safety alignment in multilingual multimodal LLMs: cross-lingual vulnerability and 'safety-by-failure'
A systematic study evaluates adversarial robustness and safety alignment of multimodal LLMs across 12 languages, finding that adversarial images optimized in one language transfer to others (cross-lingual transferability). The paper introduces the concept of 'safety-by-failure': low-resource languages appear safer not due to genuine alignment but because models fail to comprehend harmful instructions in those languages. Models like Qwen3-VL that integrate multilingual capability throughout training (rather than only at instruction tuning) show genuine cross-lingual safety with active refusal. The findings challenge the assumption that low-resource language safety metrics reflect real alignment.
LLMs fail to consistently simulate demographic perspective-taking in hate speech annotation
A new arXiv paper evaluates whether persona-conditioned LLMs can replicate how different demographic groups perceive hate speech, testing three dimensions: inter-group disagreement, in-group sensitivity, and vicarious prediction. No model consistently captures all three dimensions, and performance is highly model-dependent rather than emerging reliably from identity prompts alone. Vicarious prompting with Llama 3.1 provides the closest approximation to human disagreement patterns across demographic axes. The findings have implications for using LLMs as proxies for diverse human annotators in content moderation tasks.
Audit finds cultural translation failures and diversity collapse in LLM-adapted math word problems across 7 languages
Researchers audited how Claude Opus 4, GPT-4.1, and Gemini 2.5 Pro adapt 60 English math word problems into seven languages spanning South Asia and Italy, annotating 6,489 entity transformations. Models agreed on transformation type only 62.5% of the time and on specific substitutions in just 33.5% of cases, meaning model choice substantially shapes the cultural world students encounter. All 21 language-model combinations exhibited 'entropy collapse'—adaptations compressed rather than expanded cultural diversity—and models produced systematic regional misattributions (e.g., Bangladeshi currency for Indian Bengali students) and cross-cultural contamination (e.g., egg hunts framed as Eid activities). The study highlights that surface plausibility masks deeper corpus-level failures invisible in individual translations.
Letting Large Models Debate: The First Multilingual LLM Debate Competition
Hugging Face introduces a multilingual LLM debate competition where large language models compete against each other in structured debates. The initiative explores multi-agent interaction, argumentation quality, and cross-lingual reasoning capabilities. This represents an evaluation framework for assessing LLM persuasion, coherence, and multilingual performance in adversarial settings.
Quantifying Cross-Linguistic Effects of Syncretism on Agreement Attraction Using LLM Processing Proxies
This paper investigates why morphological syncretism amplifies agreement attraction errors in some languages (English, German, Russian) but not others (Turkish, Armenian), a pattern lacking a principled account. The authors use surprisal and attention entropy derived from large language models as proxies for human sentence processing across four languages. LLM-derived measures successfully replicate behavioral findings in English and German, align with Turkish null results, and partially capture Russian patterns. The work demonstrates LLMs as tools for cross-linguistic psycholinguistic investigation.
Study finds local languages provide better cultural knowledge access in LLMs once proficiency is controlled
A new arXiv paper introduces a controlled evaluation framework to disentangle language proficiency from culture-specific knowledge access in LLMs. Using real-world cultural questions across 13 locales and ~80 models, the authors apply item response theory to show that while English dominates on culture-agnostic questions, local languages yield a consistent knowledge-access advantage on culture-specific questions once proficiency differences are factored out. The finding challenges the common interpretation that weaker local-language accuracy implies weaker cultural knowledge, and has implications for how multilingual and regionally-aligned models are evaluated.
Tracing the Emergence of Human-Like Pragmatic Reasoning in LLMs Across Languages
Researchers conducted a population-matching experiment evaluating 25 LLMs on conditional inference tasks across four languages, comparing model behavior to matched human populations. The study finds that LLMs function as accurate semantic operators but systematically fail to capture pragmatic enrichments—context-sensitive inferences beyond literal logical meaning—that humans apply effortlessly. Model performance on pragmatic reasoning is not predicted by open vs. closed weights, training orientation, or architecture type, suggesting pragmatic reasoning remains an emergent and unreliable capability. The findings contribute to ongoing debates about whether LLMs reason like humans or merely approximate surface-level linguistic patterns.