SBSFigures: Pre-training Figure QA from Synthesized Images

A dataset for figure question-answering, synthesized for pre-training models. It was created by researchers including Risa Shionoda and Kuniaki Saito for the AAAI-25 Workshop on Document Understanding and Intelligence. The dataset page was last updated on 2026-01-18.

Use Cases

• Pre-training vision-language models based on synthesized figure images.
• Evaluating figure question-answering capabilities based on extracted attributes like text and colors.
• Benchmarking document understanding systems on a dedicated figure QA task.

Strengths

• Dataset is associated with a peer-reviewed workshop paper, suggesting a research foundation.
• The synthesis approach aims to address the challenge of gathering large-scale, annotated figure data.

Limitations

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

Provenance

Source

omron-sinicx on Hugging Face

Collection Method

Stage-by-stage synthesized images, as described in the associated paper.

Freshness

Last updated 2026-01-18 22:36:15; freshness should be verified.