Explain or connect
Improve a teaching note, add a literature pointer, identify a missing assumption, or connect an existing declaration to a textbook theorem.
Contribute to BanditRLlib
BanditRLlib welcomes sourced mathematics from bandits, reinforcement learning, probability, optimization, statistics, and adjacent fields. Every proposal enters the existing ABRL hierarchy, keeping provenance, meaning, Lean text, dependencies, obligations, and verification status together.
Improve a teaching note, add a literature pointer, identify a missing assumption, or connect an existing declaration to a textbook theorem.
Supply a sourced mathematical statement, LaTeX, a Lean draft or signature, expected imports, and dependencies. A draft is visibly marked proposed.
Contribute a compiling Lean implementation with tests and teaching prose. Maintainers still review API placement, assumptions, attribution, and integration.
A GitHub issue is enough to begin. Large formalizations should agree on scope and module ownership before substantial proof work. No proposal is displayed as compiled merely because it contains Lean-looking text.
flowchart LR
Proposal["Researcher proposal<br/>theorem + assumptions + source"] --> Packet["Contribution packet<br/>schema 1.1"]
Packet --> Middle["ABRL middle layer<br/>route and dependency review"]
Middle --> Lower["ABRL lower layer<br/>Lean implementation"]
Lower --> Compiler["Lean 4 compilation"]
Compiler --> Review["Statement fence + reviewer + full gate"]
Review -->|changes requested| Proposal
Review -->|accepted| Library["BanditRLlib on main"]
Library --> Credit["Contributor registry and release notes"]
main, and included in a verified snapshot.The community unit is a versioned JSON lemma packet. Live Formalization exports it today, including BanditRLlib reuse, Mathlib/LML candidates, semantic status, compiler status, and unresolved obligations. The packet becomes an ABRL task rather than a second proof system.
{
"schema_version": "1.1",
"id": "domain-short-lemma-name",
"status": "proposed",
"mathematics": { "plain": "...", "latex": "..." },
"lean": { "imports": ["BanditRLProof"], "code": "...", "banditrl_reused": [] },
"provenance": { "source": "DOI, arXiv, book, or original" },
"contributor": { "name": "...", "credit": "..." },
"unresolved_proof_obligations": ["semantic review"]
}
Loading the public registry…
Contributor name, preferred credit, source provenance, and review history remain in the packet and the eventual teaching note.
Review may strengthen implementation details, but it must not silently weaken the mathematical target or hide a missing proof behind a theorem card.
Public proposals are open; core-library integration follows mathematical review, namespace and API review, license checks, and the repository's Lean gate.