arXiv Author Affiliations Matched to ROR Identifiers from 2025

arXiv author affiliation data extracted from all works as of December 2025, processed via PDF-to-markdown conversion and a specialized parsing model. The extracted affiliations are matched to Research Organization Registry (ROR) identifiers. The dataset was created by cometadata and last updated on January 14, 2026.

Use Cases

• Map institutional collaboration networks based on matched ROR IDs.
• Analyze geographic or organizational trends in arXiv submissions based on affiliation data.
• Validate or improve automated affiliation parsing and entity linking models.
• Study the distribution of research output across universities and labs.

Strengths

• Covers all arXiv works as of a specific date (December 2025).
• Affiliations are matched to standardized ROR identifiers.
• Extraction performed using a named, specialized model (cometadata/affiliation-parsing-lora-Qwen3-8B-distil-GLM_4.5_Air).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect temporal and source bias inherent to the arXiv preprint platform.

Provenance

Source

arXiv preprint repository.

Collection Method

PDFs converted to markdown, then affiliations extracted and matched to ROR IDs using specified models.

Time Range

Covers works up to December 2025.

Freshness

Last updated 2026-01-14 00:07:00; freshness should be verified.

Geography

Global, based on international arXiv submissions.