MQUD: 1,250 Multimodal Questions Grounded in Scientific Figures

MQUD contains 1,250 figure-grounded inquisitive questions sourced from scientific papers. Each example pairs a scientific figure with paper context, a question, an extractive answer, and question type. The dataset was created by author 'lingchensanwen' and was last updated on Hugging Face in May 2026.

Use Cases

• Train models for visual question answering based on scientific figures and textual context.
• Benchmark machine reading comprehension systems on multimodal scientific content.
• Analyze question types and reasoning patterns in scientific inquiry.
• Develop extractive answer generation models for figure-based queries.

Strengths

• Contains 1,250 annotated multimodal question-answer pairs.
• Each entry includes a scientific figure, paper context, question, answer, and question type.
• Data is available in both JSONL and Parquet formats.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known (1,250), but the total size and sample diversity are not detailed.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Scientific papers, aggregated by author 'lingchensanwen' on Hugging Face.

Collection Method

Likely manually or semi-automatically curated from scientific publications.

Freshness

Last updated 2026-05-18 13:14:02; freshness should be verified.