Resources
Official NativQA resource hub for the framework, datasets, metrics, papers, and benchmark pages for multilingual and multimodal LLM evaluation.
The resources below collect the official framework links and dedicated pages for datasets and related work.
Resource pages
Dataset
MultiNativQA Dataset
Explore dataset links, live download metrics, language coverage, regional scope, and topic distribution for the multilingual natural QA benchmark.
Spoken visual QA
EverydayMMQA and OASIS
Explore the multilingual and multimodal framework for culturally grounded spoken visual QA, with the paper summary, pipeline, dataset scale, and benchmark findings in one place.
AudioLLM resource
MENASpeechBank
Explore the Arabic and MENA speech resource with a reference voice bank, persona-conditioned multi-turn conversations, and live Hugging Face download metrics.
Spoken QA dataset
SpokenNativQA
A multilingual resource of everyday spoken queries for evaluating LLMs in speech-centric, low-resource, and culturally grounded settings.
Evaluation toolkit
OmniScore
A deterministic multilingual generative text evaluation toolkit with released data and model resources for reproducible evaluation beyond LLM-as-a-judge.
NativQA Framework
Quick install pip install nativqa-framework
All-time PyPI downloads
Git clone downloads (public API)
GitLab stars
GitLab forks
The NativQA Framework provides the collection and processing pipeline used to build culturally aligned QA resources from native-speaker and region-aware queries.
It supports multilingual dataset construction workflows that can feed both benchmarking and fine-tuning, and it also serves as the foundation for the public resource pages linked above.