Tabverse: A Multimodal Benchmark for Cross-Format Table Understanding

MBZUAI and SUTD researchers created Tabverse, a controlled multimodal table benchmark. It aligns HTML, Markdown, and LaTeX table representations with rendered PNG images to evaluate table understanding in large language and vision models. The dataset was last updated on June 5, 2026.

Use Cases

• Benchmarking model performance on table parsing based on the multimodal HTML, Markdown, and LaTeX representations.
• Evaluating cross-format consistency in table understanding based on the alignment between code and image representations.
• Training models for robust table interpretation based on the controlled multimodal benchmark structure.

Strengths

• Dataset is explicitly designed as a controlled benchmark for systematic evaluation.
• Multimodal structure aligns three distinct table formats (HTML, Markdown, LaTeX) with visual PNG renderings.

Limitations

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

Provenance

Source

Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) and Singapore University of Technology and Design (SUTD).

Collection Method

Created as a research benchmark; specific data gathering method is not detailed in the provided description.

Time Range

null

Freshness

Last updated 2026-06-05 13:38:15; freshness should be verified.

Geography

null