INDOTABVQA: Cross-Lingual Table Visual Question Answering for Indonesian Documents

INDOTABVQA is a benchmark dataset for evaluating Vision-Language Models on cross-lingual table understanding in Bahasa Indonesia document images. The dataset was created by NusaBharat and is associated with a paper accepted at ACL 2026 Findings. It was last updated on the Hugging Face platform on April 9, 2026.

Use Cases

• Benchmarking cross-lingual understanding capabilities of VLMs based on the dataset's stated focus.
• Evaluating numerical and structural reasoning over tables in document images.
• Testing document-level table comprehension in a non-English language context.

Strengths

• Dataset is associated with a peer-reviewed paper accepted at ACL 2026 Findings.
• Focuses on a specific, high-value task: cross-lingual table understanding in document images.
• Last updated on 2026-04-09, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license information are unknown, which may limit suitability assessment.

Provenance

Source

NusaBharat on Hugging Face.

Collection Method

Likely created for the associated research paper; specific gathering method is not detailed.

Time Range

The dataset's temporal coverage is not specified.

Freshness

Last updated 2026-04-09 07:25:44.

Geography

Focuses on documents in Bahasa Indonesia, suggesting a primary geographic relevance to Indonesia.