Multilingual Language Model Evaluation

Sentence simplification aims to make complex text more accessible by reducing linguistic complexity while preserving the original meaning. However, progress in this area remains limited for mid-resource and low-resource languages due to the scarcity of high-quality data. To address this gap, we introduce the OasisSimp dataset, a multilingual dataset for sentence-level simplification covering five languages: English, Sinhala, Tamil, Pashto, and Thai. Among these, no prior sentence simplification datasets exist for Thai, Pashto, and Tamil, while limited data is available for Sinhala. Each language simplification dataset was created by trained annotators who followed detailed guidelines to simplify sentences while maintaining meaning, fluency, and grammatical correctness. We evaluate eight open-weight multilingual Large Language Models (LLMs) on the OasisSimp dataset and observe substantial performance disparities between high-resource and low-resource languages, highlighting the simplification challenges in multilingual settings. The OasisSimp dataset thus provides both a valuable multilingual resource and a challenging benchmark, revealing the limitations of current LLM-based simplification methods and paving the way for future research in low-resource sentence simplification. The dataset is available at https://OasisSimpDataset.github.io/.

Bibliographic reference extraction and parsing are foundational for citation indexing, linking, and downstream scholarly knowledge-graph construction. However, most established evaluations focus on clean, English, end-of-document bibliographies, and therefore underrepresent the Social Sciences and Humanities (SSH), where citations are frequently multilingual, embedded in footnotes, abbreviated, and shaped by heterogeneous historical conventions. We present a unified benchmark that targets these SSH-realistic conditions across three complementary datasets: CEX (English journal articles spanning multiple disciplines), EXCITE (German/English documents with end-section, footnote-only, and mixed regimes), and LinkedBooks (humanities references with strong stylistic variation and multilinguality). We evaluate three tasks of increasing difficulty -- reference extraction, reference parsing, and end-to-end document parsing -- under a schema-constrained setup that enables direct comparison between a strong supervised pipeline baseline (GROBID) and contemporary LLMs (DeepSeek-V3.1, Mistral-Small-3.2-24B, Gemma-3-27B-it, and Qwen3-VL (4B-32B variants)). Across datasets, extraction largely saturates beyond a moderate capability threshold, while parsing and end-to-end parsing remain the primary bottlenecks due to structured-output brittleness under noisy layouts. We further show that lightweight LoRA adaptation yields consistent gains -- especially on SSH-heavy benchmarks -- and that segmentation/pipelining can substantially improve robustness. Finally, we argue for hybrid deployment via routing: leveraging GROBID for well-structured, in-distribution PDFs while escalating multilingual and footnote-heavy documents to task-adapted LLMs.

Research on developmentally plausible language models has largely focused on English, leaving open questions about multilingual settings. We present a systematic study of compact language models by extending BabyBERTa to English-French scenarios under strictly size-matched data conditions, covering monolingual, bilingual, and cross-lingual settings. Our design contrasts two types of training corpora: (i) child-directed speech (about 2.5M tokens), following BabyBERTa and related work, and (ii) multi-domain corpora (about 10M tokens), extending the BabyLM framework to French. To enable fair evaluation, we also introduce new resources, including French versions of QAMR and QASRL, as well as English and French multi-domain corpora. We evaluate the models on both syntactic and semantic tasks and compare them with models trained on Wikipedia-only data. The results reveal context-dependent effects: training on Wikipedia consistently benefits semantic tasks, whereas child-directed speech improves grammatical judgments in monolingual settings. Bilingual pretraining yields notable gains for textual entailment, with particularly strong improvements for French. Importantly, similar patterns emerge across BabyBERTa, RoBERTa, and LTG-BERT, suggesting consistent trends across architectures.

As Large Language Models (LLMs) becomes a popular source for religious knowledge, it is important to know if it treats different groups fairly. This study is the first to measure how LLMs handle the differences between the two main sects of Islam: Sunni and Shia. We present a test called SectEval, available in both English and Hindi, consisting of 88 questions, to check the bias-ness of 15 top LLM models, both proprietary and open-weights. Our results show a major inconsistency based on language. In English, many powerful models DeepSeek-v3 and GPT-4o often favored Shia answers. However, when asked the exact same questions in Hindi, these models switched to favoring Sunni answers. This means a user could get completely different religious advice just by changing languages. We also looked at how models react to location. Advanced models Claude-3.5 changed their answers to match the user's country-giving Shia answers to a user from Iran and Sunni answers to a user from Saudi Arabia. In contrast, smaller models (especially in Hindi) ignored the user's location and stuck to a Sunni viewpoint. These findings show that AI is not neutral; its religious ``truth'' changes depending on the language you speak and the country you claim to be from. The data set is available at https://github.com/secteval/SectEval/

With the rapid advances of large language models, it becomes increasingly important to systematically evaluate their multilingual and multicultural capabilities. Previous cultural evaluation benchmarks focus mainly on basic cultural knowledge that can be encoded in linguistic form. Here, we propose SommBench, a multilingual benchmark to assess sommelier expertise, a domain deeply grounded in the senses of smell and taste. While language models learn about sensory properties exclusively through textual descriptions, SommBench tests whether this textual grounding is sufficient to emulate expert-level sensory judgment. SommBench comprises three main tasks: Wine Theory Question Answering (WTQA), Wine Feature Completion (WFC), and Food-Wine Pairing (FWP). SommBench is available in multiple languages: English, Slovak, Swedish, Finnish, German, Danish, Italian, and Spanish. This helps separate a language model's wine expertise from its language skills. The benchmark datasets were developed in close collaboration with a professional sommelier and native speakers of the respective languages, resulting in 1,024 wine theory question-answering questions, 1,000 wine feature-completion examples, and 1,000 food-wine pairing examples. We provide results for the most popular language models, including closed-weights models such as Gemini 2.5, and open-weights models, such as GPT-OSS and Qwen 3. Our results show that the most capable models perform well on wine theory question answering (up to 97% correct with a closed-weights model), yet feature completion (peaking at 65%) and food-wine pairing show (MCC ranging between 0 and 0.39) turn out to be more challenging. These results position SommBench as an interesting and challenging benchmark for evaluating the sommelier expertise of language models. The benchmark is publicly available at https://github.com/sommify/sommbench.

Recent progress in Natural Language Processing (NLP) has been driven by the emergence of Large Language Models (LLMs), which exhibit remarkable generative and reasoning capabilities. However, despite their success, evaluating the true semantic understanding of these models remains a persistent challenge. Traditional benchmarks such as Word-in-Context (WiC) effectively probe this capability, but their creation is resource-intensive and often limited to high-resource languages. In this paper, we introduce SemBench, a framework for automatically generating synthetic benchmarks that assess the semantic competence of LLMs using only dictionary sense definitions and a sentence encoder. This approach eliminates the need for curated example sentences, making it both scalable and language-independent. We evaluate SemBench in three languages (English, Spanish, and Basque) spanning different levels of linguistic resources, and across a wide range of LLMs. Our results show that rankings derived from SemBench strongly correlate with those obtained from standard WiC datasets. Furthermore, our analysis demonstrates that only a small number of examples is required to achieve stable and meaningful rankings. Overall, SemBench provides a lightweight, adaptable, and data-efficient framework for cross-lingual evaluation of semantic understanding in LLMs.

Tiny Aya redefines what a small multilingual language model can achieve. Trained on 70 languages and refined through region-aware posttraining, it delivers state-of-the-art in translation quality, strong multilingual understanding, and high-quality target-language generation, all with just 3.35B parameters. The release includes a pretrained foundation model, a globally balanced instruction-tuned variant, and three region-specialized models targeting languages from Africa, South Asia, Europe, Asia-Pacific, and West Asia. This report details the training strategy, data composition, and comprehensive evaluation framework behind Tiny Aya, and presents an alternative scaling path for multilingual AI: one centered on efficiency, balanced performance across languages, and practical deployment.

We present the Multilingual Reasoning Gym, an extension of Reasoning Gym (Stojanovski et al., 2025), that procedurally generates verifiable reasoning problems across 14 languages. We translate templates for 94 tasks with native-speaker validation in 10 languages and targeted code or template adaptations to ensure linguistic naturalness. The Multilingual Reasoning Gym preserves the core benefits of the procedural generation approach used in the original Reasoning Gym, such as virtually unlimited problem instance generation and adjustable difficulty, and remains directly usable for Reinforcement Learning from Verifiable Rewards and evaluation settings. Problems in the Multilingual Reasoning Gym are parallel across languages, enabling crosslingually parallel data generation at massive scale due to the procedural nature of the environments. We release our implementation to support research into multilingual reasoning models.

Large language models (LLMs) have become a standard for multilingual evaluation, yet they exhibit a severe systematic translationese bias. In this paper, translationese bias is characterized as LLMs systematically favoring machine-translated text over human-authored references, particularly in low-resource languages. We attribute this bias to spurious correlations with (i) latent manifold alignment with English and (ii) cross-lingual predictability. To mitigate this bias, we propose DIBJudge, a robust fine-tuning framework that learns a minimally sufficient, judgment-critical representation via variational information compression, while explicitly isolating spurious factors into the dedicated bias branch. Furthermore, we incorporate a cross-covariance penalty that explicitly suppresses statistical dependence between robust and bias representations, thereby encouraging effective disentanglement. Extensive evaluations on multilingual reward modeling benchmarks and a dedicated translationese bias evaluation suite demonstrate that the proposed DIBJudge consistently outperforms strong baselines and substantially mitigates translationese bias.

We present a configurable pipeline for generating multilingual sets of entities with specified characteristics, such as domain, geographical location and popularity, using data from Wikipedia and Wikidata. These datasets are intended for evaluating the factuality of LLMs' long-form generation, thereby complementing evaluation based on short-form QA datasets. We present the RiDiC dataset as an example of this approach. RiDiC contains 3,000 entities from three domains -- rivers, natural disasters, and car models -- spanning different popularity tiers. Each entity is accompanied by its geographical location, English and Chinese names (if available) and relevant English and Chinese Wikipedia content, which is used to evaluate LLMs' responses. Generations about RiDiC entities were obtained from three LLMs in English and Chinese. These were then evaluated using a third-party factuality checker, which showed that entities from our dataset caused even frontier models to hallucinate. To facilitate the evaluation of LLMs' long-form factuality in multiple languages, the code, data, and generation/evaluation scripts have been released.

Speech Large Language Models (SLLMs) have rapidly expanded, supporting a wide range of tasks. These models are typically evaluated using text prompts, which may not reflect real-world scenarios where users interact with speech. To address this gap, we introduce DoWhatISay (DOWIS), a multilingual dataset of human-recorded spoken and written prompts designed to pair with any existing benchmark for realistic evaluation of SLLMs under spoken instruction conditions. Spanning 9 tasks and 11 languages, it provides 10 prompt variants per task-language pair, across five styles. Using DOWIS, we benchmark state-of-the-art SLLMs, analyzing the interplay between prompt modality, style, language, and task type. Results show that text prompts consistently outperform spoken prompts, particularly for low-resource and cross-lingual settings. Only for tasks with speech output, spoken prompts do close the gap, highlighting the need for speech-based prompting in SLLM evaluation.

Validating evaluation metrics for NLG typically relies on expensive and time-consuming human annotations, which predominantly exist only for English datasets. We propose LLM as a Meta-Judge, a scalable framework that utilizes LLMs to generate synthetic evaluation datasets via controlled semantic degradation of real data, replacing human judgment. We validate our approach using meta-correlation, measuring the alignment between metric rankings derived from synthetic data and those from standard human benchmarks. Experiments across Machine Translation, Question Answering, and Summarization demonstrate that synthetic validation serves as a reliable proxy for human judgment, achieving meta-correlations exceeding 0.9 in multilingual QA and proves to be a viable alternative where human judgments are unavailable or too expensive to obtain. Our code and data are publicly available at https://github.com/eiglerl/meta-judge.

Translations often carry traces of the source language, a phenomenon known as translationese. We introduce the first freely available English-to-Swedish dataset contrasting translationese sentences with idiomatic alternatives, designed to probe intrinsic preferences of language models. It includes error tags and descriptions of the problems in the original translations. In experiments evaluating smaller Swedish and multilingual LLMs with our dataset, we find that they often favor the translationese phrasing. Human alternatives are chosen more often when the English source sentence is omitted, indicating that exposure to the source biases models toward literal translations, although even without context models often prefer the translationese variant. Our dataset and findings provide a resource and benchmark for developing models that produce more natural, idiomatic output in non-English languages.

We investigate whether performing reasoning in a continuous latent space leads to more robust multilingual capabilities. We compare Continuous Chain-of-Thought (using the CODI framework) against standard supervised fine-tuning across five typologically diverse languages: English, Chinese, German, French, and Urdu. Our experiments on GSM8k and CommonsenseQA demonstrate that continuous reasoning significantly outperforms explicit reasoning on low-resource languages, particularly in zero-shot settings where the target language was not seen during training. Additionally, this approach achieves extreme efficiency, compressing reasoning traces by approximately $29\times$ to $50\times$. These findings indicate that continuous latent representations naturally exhibit greater language invariance, offering a scalable solution for cross-lingual reasoning.

Multilingual large language models (LLMs) are increasingly used in socially sensitive mental health contexts, including support chatbots, screening, and content moderation. This raises a reliability question: do semantically equivalent mental health inputs elicit comparable evaluations across languages, or systematic shifts consistent with language-associated social and cultural contexts? We examine this question in an English-Chinese setting with GPT-4o and Qwen3-32B using a two-level framework: construct-level evaluative orientation, measured by psychometric stigma instruments, and decision-level behavior, measured by binary stigma detection and four-class depression severity classification. Across instruments and models, Chinese prompts elicit higher stigma-related scores than English prompts. At the decision level, Chinese prompts reduce sensitivity to stigmatizing content and produce more conservative depression severity judgments, leading to more under-estimation errors. These findings show that prompt language can shift both evaluative orientation and downstream behavior in LLM-based mental health evaluation. They highlight the need to evaluate multilingual LLMs not only for aggregate performance, but also for whether they apply comparable evaluative standards across languages in socially sensitive domains.

Large Language Models (LLMs) are increasingly being deployed in multilingual, multicultural settings, yet their reliance on predominantly English-centric training data risks misalignment with the diverse cultural values of different societies. In this paper, we present a comprehensive, multilingual audit of the cultural alignment of contemporary LLMs including GPT-4o-Mini, Gemini-2.5-Flash, Llama 3.2, Mistral and Gemma 3 across India, East Asia and Southeast Asia. Our study specifically focuses on the sensitive domain of religion as the prism for broader alignment. To facilitate this, we conduct a multi-faceted analysis of every LLM's internal representations, using log-probs/logits, to compare the model's opinion distributions against ground-truth public attitudes. We find that while the popular models generally align with public opinion on broad social issues, they consistently fail to accurately represent religious viewpoints, especially those of minority groups, often amplifying negative stereotypes. Lightweight interventions, such as demographic priming and native language prompting, partially mitigate but do not eliminate these cultural gaps. We further show that downstream evaluations on bias benchmarks (such as CrowS-Pairs, IndiBias, ThaiCLI, KoBBQ) reveal persistent harms and under-representation in sensitive contexts. Our findings underscore the urgent need for systematic, regionally grounded audits to ensure equitable global deployment of LLMs.

The safety evaluation of large language models (LLMs) remains largely centered on English, leaving non-English languages and culturally grounded risks underexplored. In this work, we investigate LLM safety in the context of the Thai language and culture and introduce ThaiSafetyBench, an open-source benchmark comprising 1,954 malicious prompts written in Thai. The dataset covers both general harmful prompts and attacks that are explicitly grounded in Thai cultural, social, and contextual nuances. Using ThaiSafetyBench, we evaluate 24 LLMs, with GPT-4.1 and Gemini-2.5-Pro serving as LLM-as-a-judge evaluators. Our results show that closed-source models generally demonstrate stronger safety performance than open-source counterparts, raising important concerns regarding the robustness of openly available models. Moreover, we observe a consistently higher Attack Success Rate (ASR) for Thai-specific, culturally contextualized attacks compared to general Thai-language attacks, highlighting a critical vulnerability in current safety alignment methods. To improve reproducibility and cost efficiency, we further fine-tune a DeBERTa-based harmful response classifier, which we name ThaiSafetyClassifier. The model achieves a weighted F1 score of 84.4%, matching GPT-4.1 judgments. We publicly release the fine-tuning weights and training scripts to support reproducibility. Finally, we introduce the ThaiSafetyBench leaderboard to provide continuously updated safety evaluations and encourage community participation. - ThaiSafetyBench HuggingFace Dataset: https://huggingface.co/datasets/typhoon-ai/ThaiSafetyBench - ThaiSafetyBench Github: https://github.com/trapoom555/ThaiSafetyBench - ThaiSafetyClassifier HuggingFace Model: https://huggingface.co/typhoon-ai/ThaiSafetyClassifier - ThaiSafetyBench Leaderboard: https://huggingface.co/spaces/typhoon-ai/ThaiSafetyBench-Leaderboard

As Large Language Models (LLMs) increasingly mediate global information access with the potential to shape public discourse, their alignment with universal human rights principles becomes important to ensure that these rights are abided by in high stakes AI-mediated interactions. In this paper, we evaluate how LLMs navigate trade-offs involving the Universal Declaration of Human Rights (UDHR), leveraging 1,152 synthetically generated scenarios across 24 rights articles and eight languages. Our analysis of eleven major LLMs reveals systematic biases where models: (1) accept limiting Economic, Social, and Cultural rights more often than Political and Civil rights, (2) demonstrate significant cross-linguistic variation with elevated endorsement rates of rights-limiting actions in Chinese and Hindi compared to English or Romanian, (3) show substantial susceptibility to prompt-based steering, and (4) exhibit noticeable differences between Likert and open-ended responses, highlighting critical challenges in LLM preference assessment.

As Large Language Models (LLMs) become integrated into daily life, they are increasingly used for personal queries, including Sexual and Reproductive Health (SRH), allowing users to chat anonymously without fear of judgment. However, current evaluation methods primarily focus on accuracy, often for objective queries in high-resource languages, and lack criteria to assess usability and safety, especially for low-resource languages and culturally sensitive domains like SRH. This paper introduces LLM Evaluation Framework (LEAF), that conducts assessments across multiple criteria: accuracy, language, usability gaps (including relevance, adequacy, and cultural appropriateness), and safety gaps (safety, sensitivity, and confidentiality). Using the LEAF framework, we assessed 14K SRH queries in Nepali from over 9K users. Responses were manually annotated by SRH experts according to the framework. Results revealed that only 35.1% of the responses were "proper", meaning they were accurate, adequate and had no major usability or safety related gaps. Insights include differences in performance between ChatGPT versions, such as similar accuracy but varying usability and safety aspects. This evaluation highlights significant limitations of current LLMs and underscores the need for improvement. The LEAF Framework is adaptable across domains and languages, particularly where usability and safety are critical, offering a pathway to better address sensitive topics.

Fake news undermines societal trust and decision-making across politics, economics, health, and international relations, and in extreme cases threatens human lives and societal safety. Because fake news reflects region-specific political, social, and cultural contexts and is expressed in language, evaluating the risks of large language models (LLMs) requires a multi-lingual and regional perspective. Malicious users can bypass safeguards through jailbreak attacks, inducing LLMs to generate fake news. However, no benchmark currently exists to systematically assess attack resilience across languages and regions. Here, we propose JailNewsBench, the first benchmark for evaluating LLM robustness against jailbreak-induced fake news generation. JailNewsBench spans 34 regions and 22 languages, covering 8 evaluation sub-metrics through LLM-as-a-Judge and 5 jailbreak attacks, with approximately 300k instances. Our evaluation of 9 LLMs reveals that the maximum attack success rate (ASR) reached 86.3% and the maximum harmfulness score was 3.5 out of 5. Notably, for English and U.S.-related topics, the defensive performance of typical multi-lingual LLMs was significantly lower than for other regions, highlighting substantial imbalances in safety across languages and regions. In addition, our analysis shows that coverage of fake news in existing safety datasets is limited and less well defended than major categories such as toxicity and social bias. Our dataset and code are available at https://github.com/kanekomasahiro/jail_news_bench.

We introduce a typology-aware diagnostic for multilingual masked language models that tests reliance on word order versus inflectional form. Using Universal Dependencies, we apply inference-time perturbations: full token scrambling, content-word scrambling with function words fixed, dependency-based head--dependent swaps, and sentence-level lemma substitution (+L), which lemmatizes both the context and the masked target label. We evaluate mBERT and XLM-R on English, Chinese, German, Spanish, and Russian. Full scrambling drives word-level reconstruction accuracy near zero in all languages; partial and head--dependent perturbations cause smaller but still large drops. +L has little effect in Chinese but substantially lowers accuracy in German/Spanish/Russian, and it does not mitigate the impact of scrambling. Top-5 word accuracy shows the same pattern: under full scrambling, the gold word rarely appears among the five highest-ranked reconstructions. We release code, sampling scripts, and balanced evaluation subsets; Turkish results under strict reconstruction are reported in the appendix.

Safety alignment in large language models relies predominantly on English-language training data. When harmful intent is expressed in low-resource languages, refusal mechanisms that hold in English frequently fail to activate. We introduce LSR (Linguistic Safety Robustness), the first systematic benchmark for measuring cross-lingual refusal degradation in West African languages: Yoruba, Hausa, Igbo, and Igala. LSR uses a dual-probe evaluation protocol - submitting matched English and target-language probes to the same model - and introduces Refusal Centroid Drift (RCD), a metric that quantifies how much of a model's English refusal behavior is lost when harmful intent is encoded in a target language. We evaluate Gemini 2.5 Flash across 14 culturally grounded attack probes in four harm categories. English refusal rates hold at approximately 90 percent. Across West African languages, refusal rates fall to 35-55 percent, with Igala showing the most severe degradation (RCD = 0.55). LSR is implemented in the Inspect AI evaluation framework and is available as a PR-ready contribution to the UK AISI's inspect_evals repository. A live reference implementation and the benchmark dataset are publicly available.

Transformer-based models such as BERT have significantly advanced Natural Language Processing (NLP) across many languages. However, Nepali, a low-resource language written in Devanagari script, remains relatively underexplored. This study benchmarks multilingual, Indic, Hindi, and Nepali BERT variants to evaluate their effectiveness in Nepali topic classification. Ten pre-trained models, including mBERT, XLM-R, MuRIL, DevBERT, HindiBERT, IndicBERT, and NepBERTa, were fine-tuned and tested on the balanced Nepali dataset containing 25,006 sentences across five conceptual domains and the performance was evaluated using accuracy, weighted precision, recall, F1-score, and AUROC metrics. The results reveal that Indic models, particularly MuRIL-large, achieved the highest F1-score of 90.60%, outperforming multilingual and monolingual models. NepBERTa also performed competitively with an F1-score of 88.26%. Overall, these findings establish a robust baseline for future document-level classification and broader Nepali NLP applications.

This paper presents a comprehensive evaluation framework for assessing the cultural competence of large language models (LLMs) in Persian. Existing Persian cultural benchmarks rely predominantly on multiple-choice formats and English-centric metrics that fail to capture Persian's morphological complexity and semantic nuance. Our framework introduces a Persian-specific short-answer evaluation that combines rule-based morphological normalization with a hybrid syntactic and semantic similarity module, enabling robust soft-match scoring beyond exact string overlap. Through systematic evaluation of 15 state-of-the-art open- and closed-source models across three culturally grounded Persian datasets, we demonstrate that our hybrid evaluation improves scoring consistency by +10 compared to exact-match baselines by capturing meaning that surface-level methods cannot detect. Our human evaluation further confirms that the proposed semantic similarity metric achieves higher agreement with human judgments than LLM-based judges. We publicly release our evaluation framework, providing the first standardized benchmark for measuring cultural understanding in Persian and establishing a reproducible foundation for cross-cultural LLM evaluation research.

The reliability of multilingual Large Language Model (LLM) evaluation is currently compromised by the inconsistent quality of translated benchmarks. Existing resources often suffer from semantic drift and context loss, which can lead to misleading performance metrics. In this work, we present a fully automated framework designed to address these challenges by enabling scalable, high-quality translation of datasets and benchmarks. We demonstrate that adapting test-time compute scaling strategies, specifically Universal Self-Improvement (USI) and our proposed multi-round ranking method, T-RANK, allows for significantly higher quality outputs compared to traditional pipelines. Our framework ensures that benchmarks preserve their original task structure and linguistic nuances during localization. We apply this approach to translate popular benchmarks and datasets into eight Eastern and Southern European languages (Ukrainian, Bulgarian, Slovak, Romanian, Lithuanian, Estonian, Turkish, Greek). Evaluations using both reference-based metrics and LLM-as-a-judge show that our translations surpass existing resources, resulting in more accurate downstream model assessment. We release both the framework and the improved benchmarks to facilitate robust and reproducible multilingual AI development.

Instruction-following benchmarks remain predominantly English-centric, leaving a critical evaluation gap for the hundreds of millions of Indic language speakers. We introduce IndicIFEval, a benchmark evaluating constrained generation of LLMs across 14 Indic languages using automatically verifiable, rule-based instructions. It comprises around 800 human-verified examples per language spread across two complementary subsets: IndicIFEval-Ground, translated prompts from IFEval (Zhou et al., 2023) carefully localized for Indic contexts, and IndicIFEval-Ground, synthetically generated instructions grounded in native Indic content. We conduct a comprehensive evaluation of major open-weight and proprietary models spanning both reasoning and non-reasoning models. While models maintain strong adherence to formatting constraints, they struggle significantly with lexical and cross-lingual tasks -- and despite progress in high-resource languages, instruction-following across the broader Indic family lags significantly behind English. We release IndicIFEval and its evaluation scripts to support progress on multilingual constrained generation (http://github.com/ai4bharat/IndicIFEval).

This article describes a novel language task, the Blackbird Language Matrices (BLM) task, inspired by intelligence tests, and illustrates the BLM datasets, their construction and benchmarking, and targeted experiments on chunking and systematicity. BLMs are multiple-choice problems, structured at multiple levels: within each sentence, across the input sequence, within each candidate answer. Because of their rich structure, these curated, but naturalistic datasets are key to answer some core questions about current large language models abilities: do LLMs detect linguistic objects and their properties? Do they detect and use systematic patterns across sentences? Are they more prone to linguistic or reasoning errors, and how do these interact? We show that BLMs, while challenging, can be solved at good levels of performance, in more than one language, with simple baseline models or, at better performance levels, with more tailored models. We show that their representations contain the grammatical objects and attributes relevant to solve a linguistic task. We also show that these solutions are reached by detecting systematic patterns across sentences. The paper supports the point of view that curated, structured datasets support multi-faceted investigations of properties of language and large language models. Because they present a curated, articulated structure, because they comprise both learning contexts and expected answers, and because they are partly built by hand, BLMs fall in the category of datasets that can support explainability investigations, and be useful to ask why large language models behave the way they do.

Large Language Models (LLMs) play a critical role in how humans access information. While their core use relies on comprehending written requests, our understanding of this ability is currently limited, because most benchmarks evaluate LLMs in high-resource languages predominantly spoken by Western, Educated, Industrialised, Rich, and Democratic (WEIRD) communities. The default assumption is that English is the best-performing language for LLMs, while smaller, low-resource languages are linked to less reliable outputs, even in multilingual, state-of-the-art models. To track variation in the comprehension abilities of LLMs, we prompt 3 popular models on a language comprehension task across 12 languages, representing the Indo-European, Afro-Asiatic, Turkic, Sino-Tibetan, and Japonic language families. Our results suggest that the models exhibit remarkable linguistic accuracy across typologically diverse languages, yet they fall behind human baselines in all of them, albeit to different degrees. Contrary to what was expected, English is not the best-performing language, as it was systematically outperformed by several Romance languages, even lower-resource ones. We frame the results by discussing the role of several factors that drive LLM performance, such as tokenization, language distance from Spanish and English, size of training data, and data origin in high- vs. low-resource languages and WEIRD vs. non-WEIRD communities.

We introduce BURMESE-SAN, the first holistic benchmark that systematically evaluates large language models (LLMs) for Burmese across three core NLP competencies: understanding (NLU), reasoning (NLR), and generation (NLG). BURMESE-SAN consolidates seven subtasks spanning these competencies, including Question Answering, Sentiment Analysis, Toxicity Detection, Causal Reasoning, Natural Language Inference, Abstractive Summarization, and Machine Translation, several of which were previously unavailable for Burmese. The benchmark is constructed through a rigorous native-speaker-driven process to ensure linguistic naturalness, fluency, and cultural authenticity while minimizing translation-induced artifacts. We conduct a large-scale evaluation of both open-weight and commercial LLMs to examine challenges in Burmese modeling arising from limited pretraining coverage, rich morphology, and syntactic variation. Our results show that Burmese performance depends more on architectural design, language representation, and instruction tuning than on model scale alone. In particular, Southeast Asia regional fine-tuning and newer model generations yield substantial gains. Finally, we release BURMESE-SAN as a public leaderboard to support systematic evaluation and sustained progress in Burmese and other low-resource languages. https://leaderboard.sea-lion.ai/detailed/MY

Recent advances in multilingual representation learning aim to bridge the performance gap between high- and low-resource languages, yet their ability to preserve affective meaning across languages remains underexplored, particularly for underrepresented languages like Bengali. This research addresses cross-lingual sentiment misalignment between Bengali and English by introducing a controlled benchmarking framework evaluating four multilingual transformer models on parallel Bengali-English sentence pairs, stratified by dialect, to assess their representational stability. We demonstrate that a compressed model architecture exhibits a 28.7% "Sentiment Inversion Rate," fundamentally misinterpreting positive semantics as negative (or vice versa). Consequently, we identify a cross-lingual sentiment skew that we call "Asymmetric Empathy," where models systematically dampen or artificially amplify the affective weight of Bengali text relative to its exact English counterpart. Finally, we expose a key vulnerability regarding dialectal representation: a "Modern Bias" in the regional model, which exhibits a 57% increase in alignment error when processing the formal Bengali register compared to modern colloquial text. As foundational encoders continue to serve as safety classifiers and reward models for LLM pipelines, cross-lingual reliability becomes a critical concern. We therefore advocate for the integration of "Affective Stability" metrics into future cross-lingual benchmarks to detect and penalize polarity inversions, particularly in low-resource settings.