DanRAG-Bench: Danish Multimodal Document Retrieval Across Five Sectors

DanRAG-Bench is the first Danish multimodal page-level document retrieval benchmark. It spans five sectors (energy, finance, health, legal, and municipalities) across 349 document pages and 471 verified queries. The dataset was created by Johanschmidt and last updated on Hugging Face in May 2026.

Use Cases

• Benchmarking multimodal retrieval systems based on Danish public-sector documents.
• Evaluating cross-sector performance of RAG models based on the five specified domains.
• Training retrieval models on mixed-format documents containing text, tables, and diagrams.
• Assessing the difficulty of real-world queries against verified document pages.

Strengths

• First benchmark of its kind for Danish multimodal page-level document retrieval.
• Covers 349 document pages and 471 verified queries across five distinct sectors.
• Based on real-world public-sector documents containing mixed formats.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Freshness should be verified as the last update timestamp is in the future (2026-05-14).

Provenance

Source

Johanschmidt on Hugging Face

Collection Method

Likely curated from Danish public-sector documents across five sectors.

Freshness

Last updated 2026-05-14 13:30:12

Geography

Denmark