ResearchMath-14k

# ResearchMath-14k ResearchMath-14k is a collection of **14,056 research-level mathematical problem records** extracted from papers, open-problem lists, workshop sheets, and related academic sources. Each record contains the original extracted question, a rewritten self-contained problem statement, taxonomy labels, and open-status metadata. Paper: [ResearchMath-14K: Scaling Research-Level Mathematics via Agents](https://arxiv.org/abs/2605.28003) ## Load ```python from datasets import load_dataset ds = load_dataset("amphora/ResearchMath-14k", split="train") print(ds[0]) ``` ## Construction Pipeline The paper describes a two-stage agentic pipeline: an extractor agent detects candidate open questions from source documents, then a refiner agent verifies status, assigns taxonomy labels, and rewrites each item into a self-contained problem statement.  ## Domain Coverage The corpus covers 11 top-level mathematical domain groups. The distribution below is the figure used in the accompanying paper.  | Taxonomy level 1 | Count | Share | |---|---:|---:| | Analysis, PDEs, and Dynamics | 3,197 | 22.74% | | Mathematical Physics | 2,031 | 14.45% | | Discrete Mathematics and Combinatorics | 1,897 | 13.50% | | Geometry and Topology | 1,846 | 13.13% | | Algebra and Representation Theory | 1,289 | 9.17% | | Applied and Computational Mathematics | 839 | 5.97% | | Number Theory | 806 | 5.73% | | Theoretical Computer Science | 749 | 5.33% | | Probability, Statistics, and ML | 636 | 4.52% | | Logic and Foundations | 455 | 3.24% | | Other / Cross-disciplinary | 311 | 2.21% | ## Open-Status Distribution | Open status | Count | Share | |---|---:|---:| | `open` | 8,313 | 59.14% | | `partially_solved` | 2,083 | 14.82% | | `solved` | 1,171 | 8.33% | | `unknown` | 2,489 | 17.71% | ## Difficulty and Dataset Positioning ResearchMath-14k is designed to occupy the gap between large lower-level math training datasets and small research-grade evaluation sets.  ## Fine-Tuning Signal The paper reports that filtered open-problem attempts provide useful supervision even when complete ground-truth solutions are unavailable.  ## Intended Use ResearchMath-14k is intended for work on mathematical problem understanding, research-level prompt construction, and training or evaluating models on self-contained research-problem statements. ## Citation If you use this dataset, please cite the paper: ``` @article{son2026researchmath, title={ResearchMath-14K: Scaling Research-Level Mathematics via Agents}, author={Son, Guijin and Yi, Seungyeop and Gwak, Minju and Ko, Hyunwoo and Jang, Wongi and Yu, Youngjae}, journal={arXiv preprint arXiv:2605.28003}, year={2026} } ```