Papers published by Bucket Foundation on the research-atlas graph and the canon. Each is free to read, born with a real DOI via Zenodo, and fully reproducible — every headline number pinned by a test suite. For the education-reform research corpus, see the education-atlas.
preprint · 2026-06-19CC-BY-4.0
We assemble a reconciled graph of the global public-research economy, 887,016 grants from the U.S. National Institutes of Health (NIH), the U.S. National Science Foundation (NSF), the European Commission (EC, via CORDIS), and UK Research and Innovation (UKRI), 2015-2025, in which recipient organizations are merged on ROR identifiers, investigators on ORCID, and 156,877 research outputs are linked to the grants that funded them through 285,604 OpenAlex acknowledgement edges.
Gianangelo Dichio · research-atlas working group
preprint · 2026-06-23CC-BY-4.0
A companion single-subfield study showed that SPECTER (a transformer pre-trained on the scientific-paper citation graph) beats a TF-IDF baseline at held-out citation prediction in High-Energy Physics (+15.4% relative MAP, p = 0.0005), a large win, measured on the citation-dense top-cited slice of one subfield. The natural question, the one a practitioner faces when reaching for a neural paper-recommender, is whether that advantage generalizes.
Bucket Foundation · research-atlas working group
preprint · 2026-06-24CC-BY-4.0
Paper 01 in this series characterized who gets research grants (institutional concentration), who shares the resulting papers (co-funding), and how much output accompanies a dollar (the funding→output rate). It did not ask what is, structurally, a prior question: what does each funder actually fund, and how distinctively?
Bucket Foundation · research-atlas working group
preprint · 2026-06-24CC-BY-4.0
Papers 01-03 in this series studied the funding graph without ever touching the people side: a grant's principal investigator (PI) was a name-only node with no ORCID, so funding could not be joined to careers at all. A new conservative resolver closes that gap, producing a grant_pi_person bridge that links 528,570 of the 1,740,326 PI edges (30.4%) to 59,180 distinct canonical researchers (490,839 edges carry an ORCID).
Bucket Foundation · research-atlas working group
Bucket Foundation report · 2026-10-08CC-BY-4.0 text; figures and data per file below
A first pass, the Solver Gap Engine of 2026-09-30, ranked 1,364 open problems from the formal-conjectures repository by their nearest solved problem in another file and found 31 at cosine similarity 0.8 or above. The frontier is the second pass. We embed 5,209 problem statements across seven canon branches with a small sentence encoder, call a problem's reach its highest similarity to a solved problem, and draw a frontier at the 10th percentile of solved problems' reach, 0.762 on this set. Open problems are sorted into a reach class: 268 close to known results, 863 borderline and 954 that need a new idea.
Gianangelo Dichio
read paper →$0.05 per citation in a paid work Bucket Foundation report · 2026-10-07CC-BY-4.0 text; figures MIT code output over the repository's own synthetic metrics; code MIT
A tool that helps a person answer today and leaves them no better at answering alone next week has made them dependent on it. We state a measurement protocol that separates the two outcomes. On one task set, three scores are taken: the person's unaided score Hp, the model's score A, and the person's joint score Jp with the model's answer in view. The multiplier is mp = Jp / max(Hp, A), the gain is D = Jp - max(Hp, A), and the unaided test is repeated after seven days to give retention Rp = Hp(t + 7 days) - Hp(t) and learning L = Jp(t + 7 days) - Hp(t + 7 days). A tool that raises Jp while L stays at or below zero across its interval marks dependence.
Gianangelo Dichio
read paper →$0.05 per citation in a paid work