efficient-vla-extracts

Parser-derived artifacts for the efficient-vla-wiki project.

This dataset is the data layer companion to the main repository:

• GitHub: efficient-vla-wiki
• Dataset repo: efficient-vla-extracts

What is inside

The staged upload keeps the local directory structure:

extracts/
├── meta/
│   ├── extract_build_index.jsonl
│   └── extract_build_status.json
└── parses/
    └── <paper_id>/

Each paper directory may contain:

• pdftotext.txt
• pdftotext.bbox.html
• docling/*.md
• docling/*.json
• parser-generated images and figures
• manifest.json

What this dataset is for

This dataset is used as the local parse layer for:

• paper lookup
• section and figure localization
• evidence backtracking
• query-time report generation
• wiki maintenance in efficient-vla-wiki

Source and generation

• Source papers are collected from arXiv into the local raw/ directory of the main repository.
• Parser artifacts are generated from local PDFs with:
  • pdftotext
  • pdfinfo
  • Docling
  • Marker

License note

This dataset contains parser-derived artifacts from third-party papers.

• Reuse conditions depend on the license of each source paper.
• Do not assume a single uniform redistribution license for all files.
• Check the source paper license before re-hosting or republishing extracted content.

Related files

• Main repository policy: docs/data-and-sync.md
• Local manifest: scripts/release/manifests/papers.jsonl
• Local HF config: scripts/release/manifests/huggingface_extracts.json