Datasets
Machine-readable catalog of training and evaluation datasets for payments, Lightning, fraud, and agent tool use. Our own library is first.
This catalog hosts metadata, cards, and our own short derivatives. A sample is 1 MB or smaller and only when the license is CC BY, CC BY-SA, MIT, Apache-2.0, or CC0, with attribution. ND, gated, unknown-license, and files over 25 MiB are linked, not mirrored.
A trained model is a possible later option. It would be hosted on Hugging Face, not on this site.
6 datasets. JSON: /datasets/index.json. JSONL: /datasets/index.jsonl. Schema: /schemas/dataset-card.schema.json. Bots: /bots/.
Agent Paywire · CC-BY-4.0 · open · 140 KB
Our own sources dataset. Listed first. Files already published at /dataset/library.jsonl, /dataset/library.csv, and /dataset/README.md. CC BY 4.0 covers this metadata and the summaries. Linked content belongs to its publishers. This catalog does not copy those files again.
Source. card.json. README.md.
Elliptic · CC-BY-NC-ND-4.0 · login · 697 MB
Bitcoin transaction graph labeled licit or illicit. A Kaggle account is required. About 697 MB (697459499 bytes). NoDerivatives: this site hosts no sample and no mirror. The Hugging Face copy yhoma/elliptic-bitcoin-dataset is wrongly tagged MIT. Use the original Kaggle dataset.
Source. card.json. README.md.
Edgar Alonso Lopez-Rojas · CC-BY-SA-4.0 · login · 494 MB
Synthetic mobile-money transactions for fraud detection. A Kaggle account is required. About 494 MB (493534783 bytes). The license allows a mirror with attribution and the same license. This site links only and does not host the file.
Source. card.json. README.md.
lnresearch · license unconfirmed (not verified) · open · 3.2 GB
Repository code is MIT. Zenodo record 10.5281/zenodo.4088530 says other-open. The data-file license is unconfirmed. Latest snapshot gossip-20230924.gsp.bz2 is about 3.2 GB (3203590306 bytes, 35984848 messages). data_period ends 2023-09. This site does not mirror the snapshots.
Source. card.json. README.md.
Salesforce · CC-BY-4.0 · terms-acceptance · 96 MB
Tool-calling training data. The Hugging Face dataset card says CC-BY-4.0. The JSON is about 96 MB (96116570 bytes). Access is gated: accept the terms on Hugging Face before download. This site does not mirror the files.
Source. card.json. README.md.
Sierra · MIT · open · 461 MB
Agent, tool, and user benchmark with retail, airline, and telecom domains. The repository is about 461 MB. This site does not mirror the repository. domains.json is our own short summary of those three domains, under MIT, with attribution to Sierra Research.
Source. card.json. README.md. Sample.