arXiv:2608.30065v1 Announce Type: cross Abstract: Large language models (LLMs) demonstrate strong Helpfulness, Harmlessness, and Honesty (3H) alignment in English-centric settings, but these gains tr...
The paper investigates how multilingual large language models perform when generating stories in Urdu, a low‑resource language. The authors created a corpus of 93 Urdu stories produced by GPT‑5.1, Qwen‑3‑Max, and DeepSeek‑3.1, and manually annotated errors across a nine‑label taxonomy covering linguistic, semantic, and cultural aspects. Findings reveal frequent grammatical and semantic mistakes, lack of coherence, unnatural repetition, and pervasive cultural shallowness, with few‑shot prompting failing to resolve many of these issues.
The paper introduces UrduFactBench and UrduFactQA, two hand‑annotated benchmarks for claim verification and factual consistency evaluation in Urdu, created through a multi‑stage annotation process with native speakers. It also presents UrduFactCheck, a modular fact‑checking framework that uses both monolingual and translation‑based evidence retrieval to address the scarcity of high‑quality Urdu evidence. Experiments on twelve LLMs show that translation‑augmented pipelines outperform monolingual ones, highlighting ongoing challenges for open‑source models in Urdu.
arXiv:2608. 14626v1 Announce Type: cross Abstract: Large Language Models (LLMs) have achieved substantial progress in safety alignment, yet their safety guarantees remain significantly weaker in low-resource and multilingual settings than in high-resource languages.
The paper introduces a new benchmark for Urdu‑to‑English idiomatic translation, featuring 4,000 manually verified sentence pairs in both native Perso‑Arabic script and Romanized Urdu. It evaluates multiple tasks—translation, paraphrasing, idiom span detection, and back‑translation—using various prompting strategies, and finds that state‑of‑the‑art large language models outperform traditional neural machine translation systems, especially in preserving figurative meaning. The study also highlights challenges posed by the lack of standardized orthography in Romanized Urdu, which affects consistency and idiom span detection.
arXiv:2610.37245v1 Announce Type: cross Abstract: Safety alignment research for large language models (LLMs) remains concentrated on high-resource languages, leaving open how existing alignment metho...
The paper titled "Safety Alignment Illusion: The Cross-Lingual Safety Gap in LLMs" highlights that current safety alignment training for large language models is predominantly English-centric, leading to failures in non‑English languages. It introduces INCLUDE, a multilingual benchmark with 2,604 prompts in six languages (English, Hindi, Bengali, Marathi, Tamil, and Hinglish) to measure Indian‑centric socio‑cultural biases. Evaluation of ten open‑ and closed‑source LLMs shows that Bengali models exhibit the highest bias scores among open‑source models, while English shows the lowest bias in open‑source but the highest in closed‑source models.
VakyArth is the first pragmatic benchmark for Indic languages, covering Hindi, Punjabi, Tamil, and Malayalam. It tests models on five pragmatic phenomena—deixis, speech acts, implicature, social pragmatics, and coherence—using multiple-choice questions, natural language inference, and translation tasks authored by native speakers. Evaluation of multilingual LLMs shows consistent failures on pragmatic meanings rooted in Indic linguistic and cultural conventions, with systematic differences across languages and tasks.
arXiv:2607. 20241v1 Announce Type: cross Abstract: Culturally loaded translation poses unique challenges for machine translation (MT), as meanings are deeply embedded in socio-cultural contexts beyond surface linguistic forms.
5-Dialects-BN is a new Bangla dialect benchmark that aligns Romanized transliteration with dialectal text, Standard Bangla, English, and subjectivity labels across five regional varieties. The dataset contains 6,000 manually annotated entries from Chittagong, Barisal, Noakhali, Sylhet, and Rangpur, each enriched with five aligned annotations produced and cross‑validated by native speakers and linguistics students. It supports tasks such as dialect identification, normalization, translation, subjectivity classification, and efficient fine‑tuning of multilingual LLMs.
The paper "Ghaib in Translation" investigates how large language models (LLMs) handle Urdu, a widely spoken language that is largely absent from safety evaluations. Five prominent LLMs—GPT‑4o, Claude Sonnet 4.5, Gemini 2.5 Flash, Qwen‑2.5, and Llama‑3.1—were tested on six datasets covering Nastaliq Urdu, Roman Urdu, English, and code‑switched Urdu‑English. The study found significant label instability between original‑script and English‑translation classifications, with missed‑in‑Urdu rates ranging from 2.4% to 9.9% (median 4.3%). A review of 205 papers across nine ALW/WOAH editions revealed no dedicated Urdu research, underscoring the language’s neglect in current safety research.
Safety alignment in large language models (LLMs) is largely developed in English, assuming these safeguards generalize across multilingual settings. However, this assumption remains underexplored and exposes a vulnerability in low-resource languages.