PubTables-v2: Large-Scale Full-Page and Multi-Page Table Extraction Dataset

PubTables-v2 is a new large-scale dataset designed for full-page and multi-page table extraction tasks. The dataset was created by rohanSingh969 and officially released on Hugging Face in February 2026. It is organized into three distinct collections, each containing tables in a specific context.

Use Cases

• Train models for full-page table structure recognition based on the dataset's focus on page-level context.
• Develop multi-page table extraction systems based on the dataset's specific multi-page collection.
• Benchmark table detection and recognition algorithms on a large-scale corpus of real-world documents.
• Fine-tune vision-language models for interpreting tables within their surrounding document content.

Strengths

• Dataset is explicitly designed for the complex task of multi-page table extraction.
• Organized into three distinct collections, suggesting structured variation in table contexts.
• Officially released on Hugging Face in February 2026, indicating a formal publication.

Limitations

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

Provenance

Source

huggingface

Freshness

Last updated 2026-04-13 10:21:01.