low-resource-languages

Tag

Cards List
#low-resource-languages

LusoClin: A Synthetic Clinical Note Dataset for Medical Information Retrieval in European Portuguese

arXiv cs.AI ↗ · yesterday Cached

LusoClin is a publicly released dataset of synthetic clinical notes in European Portuguese for privacy-preserving medical information retrieval research, generated by fine-tuning Llama 3.2 (3B) with QLoRA and annotated with a BERT-based clinical entity model. Evaluations show the dataset supports downstream IR tasks, though annotators can often distinguish synthetic from real notes.

0 favorites 0 likes
#low-resource-languages

Cross-cultural Adaptation of Recipes for Low-Resource Languages: A Benchmark and Multilingual Culinary Lexicon for Hausa, Yor\`ub\'a, and Igbo

arXiv cs.CL ↗ · 2d ago Cached

The paper introduces ReciLoReL, a benchmark of 600 culturally adapted recipes, and ReciLex, a ~15,000-entity multilingual culinary lexicon for Hausa, Yorùbá, and Igbo, evaluating how lexicon-augmented LLMs improve culturally adaptive recipe translation from English.

0 favorites 0 likes
#low-resource-languages

TUSI: A Citation-Grounded Graph-Based Benchmark

arXiv cs.CL ↗ · 3d ago Cached

The paper introduces TUSI-Bench, a Persian legal RAG evaluation benchmark built on a multi-layered knowledge framework of 7,261 legal articles with a supporting knowledge graph, offering TUSI-QA and TUSI-Bar tasks; graph-based retrieval improves recall by up to 8 percentage points.

0 favorites 0 likes
#low-resource-languages

VEPO: Mitigating Multilingual Hallucinations in LVLMs via Visual Evidence-guided Preference Optimization

arXiv cs.CL ↗ · 5d ago Cached

The paper proposes VEPO, a training-time visual evidence-guided preference optimization method that reduces multilingual object hallucinations in large vision-language models by transferring visual grounding from English to low-resource languages without external models.

0 favorites 0 likes
#low-resource-languages

CamSpokenNLI: A Pilot Resource for Cross-Variety Transfer Evaluation in Cameroon Pidgin English and Camfranglais

arXiv cs.CL ↗ · 5d ago Cached

The paper introduces CamSpokenNLI, a pilot spoken NLI dataset of 228 items covering Cameroon Pidgin English and Camfranglais, and evaluates XLM-RoBERTa cross-variety transfer and Whisper transcription. Results show modest accuracy, sensitivity to counterpart leakage and hypothesis-only baselines, and poor zero-shot ASR performance, highlighting the need for adapted evaluation and recognition for these varieties.

0 favorites 0 likes
#low-resource-languages

TRIDENT and SHAKTI: A Framework and Benchmark for Hindi AI-Text Detection

arXiv cs.AI ↗ · 2026-11-02 Cached

The paper introduces TRIDENT, a round-trip translation-based framework with three classification variants, and SHAKTI, a large-scale benchmark dataset for detecting AI-generated Hindi text, evaluated across multiple genres, LLMs, adversarial settings, and seven languages.

0 favorites 0 likes
#low-resource-languages

Aligning Large Language Models for Safety in African Languages: A Case Study of Hausa

arXiv cs.CL ↗ · 2026-10-28 Cached

研究了大语言模型在豪萨语等低资源语言中的安全对齐问题,提出HausaSafeLM安全偏好数据集,并发现仅靠拒绝率指标无法可靠衡量低资源语言中的模型安全性。

0 favorites 0 likes
#low-resource-languages

Text Generation in Low-Resource Languages Using Human-Curated Vocabularies

arXiv cs.CL ↗ · 2026-10-28 Cached

该论文提出一个结合人工整理词汇表与少量人类范文的框架,利用启用推理的LLM为极低资源乃至未见语言生成流畅文本,为合成数据自举提供了一条实用路径。

0 favorites 0 likes
#low-resource-languages

The State and Fate of Multilingual, Contextual Evaluation in the NLP World

arXiv cs.CL ↗ · 2026-10-28 Cached

本文对51个近期多语言基准(涵盖242个数据集、219种语言)进行了数据驱动的审计,从覆盖度、代表性和严谨性三方面揭示了多语言评估的结构性缺陷,如英语中心化翻译、语言覆盖稀薄和基准污染问题,并提出了构建原生、文化相关且抗污染的评估框架的建议。

0 favorites 0 likes
#low-resource-languages

Developing a Speech Language Model for Basque

arXiv cs.CL ↗ · 2026-10-28 Cached

Researchers present the first open Speech Language Model for Basque, based on the LLaMA-Omni architecture with a Latxa backbone, using translated and TTS-synthesized data to overcome low-resource scarcity for end-to-end spoken interaction.

0 favorites 0 likes
#low-resource-languages

Merge-Based Vocabulary Extension Stalls at the Pre-Tokenizer: An English-Preserving Repair for Nepali, Tested in Continued Pretraining

arXiv cs.CL ↗ · 2026-10-27 Cached

The paper shows that letters-only pre-tokenizer regexes in Llama-3 and Qwen3 impose an irreducible ~2.2x token premium on Nepali, and proposes an English-preserving regex repair that, with 32k new merges, reduces the premium to 1.01x, testing it via continued pretraining experiments.

0 favorites 0 likes
#low-resource-languages

Whispering Wisely: Smart Subset Selection for ASR Fine-Tuning

arXiv cs.CL ↗ · 2026-10-27 Cached

This paper introduces an embedding-based subset selection method using Whisper's pretrained encoder and a greedy k-center algorithm to pick diverse training subsets for ASR fine-tuning. Evaluated on Hebrew and Danish, it shows k-center selection beats random sampling, especially at small subset sizes, due to greater lexical diversity.

0 favorites 0 likes
#low-resource-languages

Is Refusal Direction Universal across Low-Resourced Languages?

arXiv cs.CL ↗ · 2026-10-22 Cached

The paper extends PolyRefuse to three additional low-resourced languages (Belarusian, Bashkir, Tajik) on Qwen2.5-14B-Instruct, finding that refusal direction transfer is graded rather than universal: directions from safety-aligned languages transfer broadly, while those from strongly misaligned languages fail to reliably modulate refusal behavior.

0 favorites 0 likes
#low-resource-languages

Global Security Blind Spots: Translation as a Safety Bypass in Low-Resource Languages

arXiv cs.CL ↗ · 2026-10-21 Cached

该论文提出一个翻译绕过威胁模型,在79种语言上评估了十个主流大模型,发现即使提示在英文下被安全拒绝,翻译成低资源语言后仍有21.9%至44.4%的绕过率,表明英文安全对齐无法可靠迁移到其他语言。

0 favorites 0 likes
#low-resource-languages

When Does Hate Transfer? A Large-Scale Study of Cross-Lingual Hate Speech Detection

arXiv cs.CL ↗ · 2026-10-21 Cached

一项覆盖14种语言的大规模研究,探讨跨语言仇恨言论检测的迁移效果,发现语言相似性是最强预测因素,而文化线索与交际规范可部分解释剩余差异。

0 favorites 0 likes
#low-resource-languages

African Languages AI Safety Readiness Report

arXiv cs.CL ↗ · 2026-10-21 Cached

A report examining the readiness of African languages for inclusion in AI safety research, classifying them into four readiness tiers based on data availability, national AI policy support, benchmark performance, and speaker population. It aims to guide investment and benchmark inclusion for underrepresented languages.

0 favorites 0 likes
#low-resource-languages

LoWAgent: A Low-Resource Multilingual Benchmark for Web Agents

arXiv cs.CL ↗ · 2026-10-21 Cached

The paper introduces LoWAgent, a low-resource multilingual benchmark evaluating web agents across English, Hindi, Bengali, Urdu, and Amharic on six localized websites with 3,000 tasks, finding notable performance drops for low-resource languages—especially open-source models—and showing that post-training on benchmark traces improves agentic performance.

0 favorites 0 likes
#low-resource-languages

Evaluating Tibetan Speech Recognition Across Dialects and Adaptation Regimes: A Multi-Granularity Analysis

arXiv cs.CL ↗ · 2026-10-21 Cached

该论文提出一个多粒度评估框架,结合字符错误率、音节错误率和分词错误率,评估藏语语音识别在多方言与不同适配策略下的表现,发现低字符错误率并不总带来更优的高层级指标,且 LoRA 适配在测试范围内仍逊于全量微调。

0 favorites 0 likes
#low-resource-languages

CulturalRiddles: A Multicultural Benchmark of Riddles

Hugging Face Daily Papers ↗ · 2026-10-21 Cached

We propose CulturalRiddles, a benchmark of 62 cultures (45 countries, 51 languages) of human-created, culturally grounded riddles requiring abductive and compositional reasoning. Even frontier LLMs reach at most 22% accuracy on some communities, highlighting gaps in cultural understanding.

0 favorites 0 likes
#low-resource-languages

@OpenAI: Our latest model advances bring stronger capabilities for building and migrating software in low-resource programming l…

X AI KOLs ↗ · 2026-10-19 Cached

OpenAI announced model advances that significantly improve coding in low-resource programming languages like COBOL, Fortran, and Visual Basic, introducing MultiLangBench and a COBOL migration evaluation where GPT-6 Astra and GPT-6.1 Sol show major accuracy and cost gains over GPT-5.6 Sol.

0 favorites 0 likes
Next →
← Back to home

Submit Feedback