NayanaBench: Multilingual Visual Question Answering Dataset Across 22 Languages

NayanaBench is a multilingual visual question answering dataset designed for evaluating multimodal AI systems. It includes 200 examples each for 22 languages, combining optical character recognition and layout analysis. The dataset was created by Nayana-cognitivelab and was last updated on July 28, 2025.

Use Cases

• Benchmarking multilingual visual question answering models based on the 22 supported languages.
• Training OCR-integrated multimodal systems based on the described optical character recognition capabilities.
• Evaluating layout understanding in AI systems based on the described layout analysis features.
• Developing AI for Indian language support based on the mentioned Kannada and Hindi examples.

Strengths

• Covers 22 distinct languages, providing linguistic diversity.
• Includes 200 examples per language, offering a consistent scale for comparison.
• Integrates multiple capabilities: visual question answering, OCR, and layout analysis.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Collection Method

Likely compiled for multimodal AI evaluation, but specific collection method is not detailed.

Freshness

Last updated 2025-07-28 09:40:52; freshness should be verified.

Geography

The dataset includes Indian languages like Kannada and Hindi, suggesting a focus on India, but the full geographic scope is not specified.