ResearchMath-14k: 14,056 Research-Level Mathematical Problem Statements

14,056 records of research-level mathematical problems extracted from academic papers, open-problem lists, and workshop sheets. Each record contains an original question, a rewritten self-contained statement, taxonomy labels, and open-status metadata. The dataset was created by 'amphora' and was last updated on the Hugging Face platform in May 2026.

Use Cases

• Training or evaluating mathematical reasoning agents based on research-level problem statements.
• Classifying mathematical problems by topic based on the provided taxonomy labels.
• Studying the structure of open research problems based on the open-status metadata.
• Generating self-contained problem statements based on the rewritten versions of extracted questions.

Strengths

• Contains 14,056 distinct research-level mathematical problem records.
• Each problem includes both an original extracted question and a rewritten self-contained statement.
• Provides structured metadata including taxonomy labels and open-status information.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but the specific data format and file size are unknown.
• Data may reflect the temporal and source bias inherent to the academic papers and lists from which it was extracted.

Provenance

Source

Extracted from academic papers, open-problem lists, and workshop sheets.

Collection Method

Extraction and rewriting process described in the associated paper 'ResearchMath-14k: Leveraging Internet-Search Agents to Scale Research-Level Mathematical Reasoning'.

Freshness

Last updated 2026-05-28 02:11:42.