PanAtlas – Kentucky Open Science

Matching: shows which people, models, and data fit each challenge
Benchmark register: scored benchmarks and openly published gaps
Evaluation records: approved evaluations from the Commons are considered for PanAtlas

Home / Platform / PanAtlas

PanAtlas is the distributed atlas: a versioned record of known capabilities, and a distributed system that runs directly on Cresco, where every PanAtlas instance, wherever it is hosted, is a node on the mesh. It sits beside the stack, connected in both directions to the modular toolkit and to Astral: the evaluations, datasets, and model results that the toolkit and the agents produce become new records in PanAtlas, and PanAtlas provides what is known (datasets, models, benchmarks, and their links) back to them. An instance can run at any participating site and joins the mesh like any other resource, with GaiaKeep as its durable, versioned backend.

Registers datasets with their provenance and per-document licensing
Links each record to entities through Wikipedia, Wikidata, the Medical Subject Headings, and the graphs of NSF’s Open Knowledge Network
Links datasets, models, and evaluations to one another through foundation-model embeddings, so clusters of shared interest form and teams can be suggested