UrduMMLU: New Benchmark Reveals Gaps in Large Language Models' Urdu Language Understanding
Researchers have introduced UrduMMLU, a 26,431-question multiple-choice benchmark spanning 26 subjects designed to evaluate large language models on Urdu language understanding. The benchmark was built from native Urdu educational sources rather than translations, covering both standard academic subjects and region-specific content. The findings reveal significant gaps in current LLMs' Urdu knowledge, particularly for culturally and regionally grounded humanities content.
A team of researchers has released UrduMMLU, a large-scale evaluation benchmark for Urdu — a language spoken by more than 230 million people — that fills a gap left by the absence of native MMLU-style assessments for the language. The benchmark comprises 26,431 multiple-choice questions across 26 subjects and five domains, sourced from native Urdu MCQ banks and public examination PDFs, with exam-derived items verified through dual human annotation and strict consensus filtering. Thirty large language models were evaluated under both English and Urdu prompts in zero-shot settings, with four open-source models additionally tested under few-shot conditions. Gemini-3.5-Flash achieved the highest accuracy at approximately 90%, while no other model surpassed 85%, and the best open-source model lagged by roughly 8 to 9 percentage points. A particularly notable finding is that many models suffer accuracy drops of 25 to 40 percentage points on Urdu-centered humanities subjects compared to STEM topics, and few-shot prompting provided only modest improvements. The authors conclude that current LLMs exhibit uneven and often inadequate Urdu knowledge, especially for regionally grounded content, underscoring the need for more representative multilingual benchmarks.
What's missing
It is unclear whether the benchmark accounts for dialectal variation within Urdu (e.g., differences between Pakistani and Indian Urdu usage). The paper is a preprint submitted to ARR May 2026 and has not yet undergone formal peer review, so findings should be treated as preliminary.
What different sources said
- arXiv cs.CLCenter
Phun-Bench: Evaluating LLMs on Phonological Understanding in Chinese
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.