// tengoku
Tengoku (天国) is one self-contained Lean 4 library. It was seeded from Mathlib and every package Mathlib pulls in, and it grows by verified translations of theorems from every open formalization project we can find, and by direct contributions. Its purpose is continuously improving, reliable context for automated theorem provers: one library to search, one toolchain, one definition of “trusted”.
This page is the full breakdown. The repository README is the quick start.
https://barkingtree-leak-i.hf.space/sse. No auth configuration required. The index is managed by CompeteMath and follows every merge of the tree.GET https://competemath.com/api/tengoku/search?q=… returns JSON, no key. Input and output shapes: docs/api.md.Tengoku/ folder. The only external requirement is the pinned Lean toolchain (v4.34.0-rc2).SEED.md tracks their origins, but we no longer sync with them. Original declaration names are unchanged.sorry), and no extra axioms.Tengoku/<Library>/. Every module is built.scripts/cache.sh get to download the latest cache and only build your changes, or use scripts/pin.sh to sync perfectly and build nothing.main in minutes without compiling.Every record already has a real proof from somewhere. The tier says whether Leak has stamped it:
tentative — a real proof from a real source (every record carries a source_url pointing at it) that Leak has not re-verified on the pinned toolchain yet.staging — a translation of a real proof, produced by Emissary-Archangel: the original lived on another toolchain; this is the same claim restated and re-proved on Tengoku’s, and it passed two gates — it compiles cleanly, and an independent entailment check confirms it proves at least as much as the original. Not yet in any module.trusted — compiled and certified by the pinned toolchain’s own kernel, in this tree. Mathlib lives here by definition.Promotion only goes one way, tentative or staging → trusted, and only by Leak actually building the record’s module. Nothing is trusted by assumption.
Each line of every data/**/*.jsonl file is one record:
{
"name": "...",
"statement": "theorem ... : ...",
"proof": ":= by ...",
"status": "tentative | staging | trusted",
"library": "...",
"source_url": "...",
"toolchain": "..."
}Every change is a pull request, and the repository does the checking:
sorry,axiom, native_decide, macros, notation or anything that runs code; a signed-off commit; no authorship line ever removed.Details, commands and the record files to edit: CONTRIBUTING.md, how Tengoku is tested, the blind re-proof test, top-ups.
Nearly every serious formalization project depends on Mathlib and tracks a Mathlib-compatible toolchain, so the reachable union of theorems is largest on the Mathlib release most projects have already moved to, not on a novel toolchain. When the tree was created, Mathlib’s master and its dependencies were pinned to v4.34.0-rc2, and the largest live projects (FLT, Carleson, PFR) matched it exactly. Everything on another toolchain is reached by translation.
100+ source libraries, 500,000+ theorems. A declaration whose proof contains sorry anywhere is dropped at harvest, never mislabelled as tentative.
| Source | Toolchain | Statements |
|---|---|---|
| leanprover-community/mathlib4 — the community mathematics library itself | v4.34.0-rc2 | 188,989 |
| Prove2Me — collaborative Lean formalization platform; harvested via its API, each record links to the theorem's own page | mixed (per record) | 54,457 |
| anthropics/fermats-last-theorem — Claude's complete, independent Lean 4 formalization of Fermat's Last Theorem (Sept 2026) | v4.33.1 | 48,501 |
| CBirkbeck/AINTLIB — a monorepo aggregating several of Chris Birkbeck's own number-theory projects; only the content unique to it (ModularCurves, FltRegularBernoulli, HasseWeil, DedekindResidue, NagellLutz) is included here — the rest duplicates the standalone repos already listed | v4.34.0-rc2 | 30,131 |
| Verified-zkEVM/evm-asm — verified EVM assembly semantics | v4.33.0 | 29,092 |
| facebookresearch/atlas-lean — Meta/FAIR's Atlas project | v4.29.0 | 20,335 |
| frenzymath/Poincare-Conjecture — background lemmas toward Perelman's proof, organized by source textbook (do Carmo, Lee, Hatcher, Evans, Gilbarg–Trudinger, Chow–Knopf, Topping, Morgan–Tian, and Kleiner–Lott's exposition of Perelman's argument) | v4.32.1 | 18,748 |
| teorth/equational_theories — Terence Tao's project mapping relations between equational theories of magmas | v4.29.1 | 13,193 |
| ColinBundschu/mass-gap — Yang-Mills mass gap over compact simple groups | v4.32.2 | 11,043 |
| CBirkbeck/LeanBridge — modular forms / LMFDB q-expansion computations, staged for eventual upstreaming to Mathlib (`ForMathlib/`) | v4.31.0 | 10,980 |
| leanprover-community/physlib — physics results (classical mechanics, QFT, quantum info) | v4.33.0 | 10,571 |
| LionSR/TNLean — tensor-network theory | v4.34.0-rc1 | 10,383 |
| siqiliu-tsinghua/tautology — the reals constructed and formalized from nothing, without Mathlib | v4.30.0 | 9,658 |
| Robby955/FormalSLT — statistical learning theory (concentration, generalization bounds) | v4.32.2 | 8,583 |
| AlexKontorovich/PrimeNumberTheoremAnd — the Prime Number Theorem and related results | v4.32.2 | 8,028 |
| Verified-zkEVM/VCVio — machine-checked cryptographic proofs | v4.33.1 | 6,621 |
| dwrensha/compfiles — catalog of competition problems formalized in Lean | v4.34.0-rc1 | 6,042 |
| CBirkbeck/uniform-sheafy-tate-domains-lean — adic spaces / uniform sheafy Tate domains (supersedes and extends Adic-Spaces below) | v4.33.0 | 5,801 |
| FormalizedFormalLogic/Foundation — first/second-order logic completeness, Gödel's incompleteness theorems, modal logic | v4.33.1 | 5,559 |
| Verified-zkEVM/CompPoly — computable polynomial arithmetic | v4.33.1 | 4,899 |
| frenzymath/FormalPantheon — three named results: bounded gaps between primes, "period three implies chaos", and Waring's problem | v4.32.0 | 4,735 |
| Verified-zkEVM/ArkLib — formally verified arguments-of-knowledge (SNARK-adjacent) library | v4.33.1 | 4,562 |
| Verified-zkEVM/PolyFun — polynomial functional representations | v4.33.1 | 4,029 |
| schildep/verified-3d-mesh-intersection — formally verified 3D mesh intersection (CSG) algorithm correctness | v4.15.0 | 3,753 |
| CBirkbeck/LeanModularForms — modular forms | v4.31.0-rc2 | 3,698 |
| formal-applied-math/formal-mathfin — formally verified mathematical finance (Black-Scholes, FTAP, Merton) | v4.32.0 | 2,885 |
| math-inc/Sphere-Packing-Lean — sphere packing optimality in dimensions 8 and 24 (Viazovska et al.) | v4.28.0 | 2,818 |
| leanprover-community/con-nf — consistency of Quine's New Foundations set theory | v4.21.0-rc3 | 2,785 |
| google-deepmind/formal-conjectures — DeepMind's formalized-conjectures benchmark (Erdős problems, Ben Green's 100 open problems, etc.); only the already-proven subset harvests here | v4.33.1 | 2,584 |
| fpvandoorn/Carleson — Carleson's theorem on pointwise convergence of Fourier series | v4.34.0-rc2 | 2,510 |
| mortarsanjaya/IMOSLLean4 — IMO Shortlist problems (2006+, all categories except Geometry) | v4.34.0-rc2 | 2,330 |
| CBirkbeck/CertifyingInvariantsNF — certifying ideal-arithmetic invariants of number fields | v4.30.0-rc1 | 2,257 |
| Lean-MoDS/StatsMLlib — probability/statistics/ML (concentration, empirical processes, random matrices) | v4.33.0 | 2,218 |
| ImperialCollegeLondon/FLT — Kevin Buzzard et al.'s formalization of Fermat's Last Theorem (ongoing; lemmas proven so far) | v4.34.0-rc2 | 2,198 |
| sinhp/HoTTLean — sorry-free groupoid model of homotopy type theory | v4.25.0-rc2 | 2,122 |
| leanprover-community/batteries — the community standard library (Mathlib's own foundation) | v4.34.0-rc2 | 1,960 |
| lecopivo/SciLean — scientific computing (differential equations, automatic differentiation) | v4.28.0-rc1 | 1,833 |
| Verified-zkEVM/clean — zkVM circuit correctness | v4.33.1 | 1,735 |
| YuanheZ/lean-stat-learning-theory — sorry-free statistical learning theory library | v4.32.0 | 1,663 |
| teorth/IEANTN — Terence Tao's analytic number theory notes/solutions | v4.34.0-rc2 | 1,522 |
| teorth/analysis — Terence Tao's Lean companion to his "Analysis I" textbook | v4.29.0-rc8 | 1,481 |
| RemyDegenne/brownian-motion — construction of Brownian motion, Kolmogorov–Chentsov continuity | v4.33.0-rc1 | 1,441 |
| EvolvingPrograms/erdos-simonovits-degeneracy — machine-checked disproof of the Erdős–Simonovits degeneracy conjecture | v4.32.0 | 1,435 |
| CBirkbeck/padic-L-functions — p-adic L-functions | v4.31.0-rc1 | 1,414 |
| urikol/QuantumOptimization — machine-verified quantum-optimization formalizations | v4.28.0 | 1,400 |
| Verified-zkEVM/riscv-zkvm — RISC-V zkVM verification | v4.33.0 | 1,396 |
| scottnarmstrong/DeGiorgi — De Giorgi–Nash–Moser elliptic PDE regularity theory | v4.29.0-rc6 | 1,192 |
| uda-lab/leray-hopf — Leray–Hopf weak solutions for Navier–Stokes | v4.31.0-rc2 | 1,151 |
| math-inc/strongpnt — the strong Prime Number Theorem, AI-formalized | v4.21.0 | 1,077 |
| CBirkbeck/TauCeti — the actual source of the library previously found only vendored (partially) inside `ai-safety-formalization-atlas`, now properly attributed | v4.31.0 | 1,022 |
| schildep/verified-polygon-intersection — formally verified polygon intersection algorithm correctness | v4.15.0 | 942 |
| thefundamentaltheor3m/Sphere-Packing-Lean — sphere packing optimality in dimension 8 (the original project) | v4.32.0 | 923 |
| teorth/pfr — Terence Tao, Yaël Dillies & Bhavik Mehta's formalization of the Polynomial Freiman-Ruzsa conjecture | v4.34.0-rc2 | 921 |
| leanprover-community/sphere-eversion — existence of sphere eversions | v4.33.0 | 892 |
| VTrelat/ZFLean — practical framework for set-theoretic development | v4.33.0 | 890 |
| mbrcic/ai-safety-formalization-atlas — AI-safety-relevant claims formalized as Lean proofs | v4.33.0 | 715 |
| logical-intelligence/erdos-unit-distance — a second, independent formalization of Alpöge's disproof of the Erdős unit-distance conjecture | v4.29.1 | 706 |
| teorth/sendov — work toward Sendov's conjecture | v4.34.0-rc1 | 678 |
| Zetetic-Dhruv/formal-learning-theory-kernel — a kernel for synthetic formalization of statistical learning theory | v4.29.0-rc6 | 651 |
| YaelDillies/apap — "Arithmetic Progressions - Almost Periodicity" (Kelley-Meka bound on Roth numbers) | v4.34.0-rc2 | 622 |
| CBirkbeck/ModularForms_Lean4 — modular forms (earlier project, predates LeanModularForms) | v4.5.0-rc1 | 579 |
| CBirkbeck/LocalClassFieldTheory — local class field theory | v4.7.0-rc2 | 556 |
| trishullab/PutnamBench — Putnam Competition (1962–2025) formalizations, Lean 4 subset | v4.27.0 | 524 |
| Ivan-Sergeyev/seymour — Seymour's decomposition theorem for regular matroids | v4.18.0 | 468 |
| project-numina/LeanGeo — geometry competition problems | v4.15.0 | 395 |
| WuProver/lean_characteristic_set — characteristic sets in algebraic geometry | v4.29.0-rc6 | 394 |
| djvelleman/HTPILeanPackage — tactics and exercises for "How to Prove It" | v4.33.0 | 384 |
| Shreyas4991/Algolean — algorithms & complexity library | v4.33.0 | 364 |
| ByteDance-Seed/Seed-Prover — 5 of the 2025 IMO problems plus an Erdős-problem lemma battery; its vendored miniCTX-v2 copies of other libraries are excluded | v4.14.0 | 347 |
| math-inc/FrontierMathOpen-Hypergraphs — an Epoch AI FrontierMath hypergraph/Ramsey-theoretic problem | v4.28.0 | 340 |
| WuProver/groebner_proj — Gröbner basis theory | v4.34.0-rc2 | 327 |
| YaelDillies/ClassFieldTheory — 2025 Clay Summer School class field theory repo | v4.25.0-rc2 | 327 |
| jsm28/AperiodicMonotilesLean — the aperiodic "hat tile"/einstein monotile | v4.34.0-rc2 | 311 |
| YaelDillies/mean-fourier | v4.34.0-rc2 | 287 |
| mo271/FormalBook — formalizing "Proofs from THE BOOK" (Aigner–Ziegler) | v4.34.0-rc2 | 279 |
| fpvandoorn/LeanCourse24 — Floris van Doorn's Bonn Lean course, winter 2024/25 | v4.13.0-rc3 | 269 |
| frenzymath/Anderson-Conjecture | v4.29.0-rc8 | 263 |
| CompeteMath's own certified problems — proofs Leak certified on this site | mixed (per record) | 262 |
| WuProver/MonomialOrderedPolynomial — monomial orderings for polynomial rings | v4.29.0-rc8 | 238 |
| ctchou/AutomataTheory — finite automata theory | v4.24.0-rc1 | 231 |
| frenzymath/Archon-FirstProof-Results — results from frenzymath's Archon autoformalization agent | v4.28.0 | 222 |
| teorth/estimate_tools — Terence Tao's asymptotic-estimate tooling | v4.20.0-rc5 | 221 |
| ProofOfKeags/btc-verified — verified Bitcoin protocol components | v4.30.0-rc2 | 218 |
| loganrjmurphy/LeanEuclid — autoformalization benchmark for Euclidean geometry | v4.19.0 | 206 |
| leanprover-community/flt-regular — Kummer's proof of FLT for regular primes | v4.34.0-rc2 | 195 |
| teorth/expdb — analytic number theory exponent database | v4.32.0 | 194 |
| b-mehta/ABC-Exceptions — constructions of exceptions to the ABC conjecture | v4.21.0-rc3 | 192 |
| teorth/symmetric_project | v4.2.0-rc1 | 188 |
| AnandGokhale/LeanForControl — database of control-theory proofs | v4.30.0-rc2 | 184 |
| leanprover-community/add-combi — additive-combinatorics sublibrary | v4.34.0-rc2 | 176 |
| a2435191/lean-logic-formalization — formalization of van den Dries's "Logic Notes" | v4.20.0-rc5 | 174 |
| verified-optimization/CvxLean — convex optimization modeling with verified correctness | v4.8.0-rc1 | 150 |
| YaelDillies/misc-yd — Yaël Dillies's miscellaneous results | v4.34.0-rc2 | 138 |
| emilyriehl/infinity-cosmos — basic formal theory of ∞-cosmoi | v4.34.0-rc1 | 134 |
| math-inc/Erdos1196 — Erdős Problem #1196 (bound on ∑1/(a·log a) for primitive sets) | v4.30.0-rc1 | 122 |
| CBirkbeck/DirichletNonvanishing — non-vanishing of Dirichlet L-functions on Re(s)=1 | v4.13.0-rc3 | 118 |
| ImperialCollegeLondon/formalising-mathematics-2024 — Kevin Buzzard's Lean 4 undergraduate course, 2024 | v4.5.0-rc1 | 111 |
| frenzymath/Archon-FirstProof-problem6-augmentation | v4.28.0 | 108 |
| frenzymath/qrcp-bounded-coherence-obstruction — bounded-coherence obstruction results | v4.30.0-rc2 | 103 |
| kim-em/erdos-unit-distance — Kim Morrison's formalization of Alpöge's disproof of the Erdős unit-distance conjecture | v4.32.2 | 102 |
| harmonic-ai/IMO2025 — Harmonic's Aristotle system, 5 of 6 IMO 2025 problems | v4.20.0-rc5 | 95 |
| YaelDillies/toric — toric varieties | v4.34.0-rc2 | 95 |
| YaelDillies/gibbs-measure | v4.34.0-rc2 | 94 |
| CBirkbeck/ModFormDims — dimension formulas for modular forms | v4.13.0-rc3 | 84 |
| lengyijun/goldbach_tm — Goldbach conjecture encoded as a 25-state Turing machine | v4.14.0-rc2 | 83 |
| Verified-zkEVM/leanerVM | v4.33.1 | 81 |
| hanwenzhu/miller-rabin — Miller–Rabin primality test correctness proof | v4.22.0 | 80 |
| vltanh/lean4-analysis-tao — formalization of Tao's "Analysis I" | v4.29.0 | 66 |
| CBirkbeck/NewtonPoly — Newton polygons | v4.28.0-rc1 | 65 |
| CBirkbeck/GLn_F_q — GL_n(F_q) representation theory | v4.8.0-rc2 | 62 |
| YaelDillies/LeanCamCombi — Cambridge combinatorics courses | v4.34.0-rc2 | 61 |
| YaelDillies/forbidden-matrix | v4.34.0-rc2 | 60 |
| lenianiva/Prismriver — music theory formalization + DSL | v4.28.0 | 52 |
| keilambda/ttfpi — "Type Theory and Formal Proof: An Introduction" | v4.13.0 | 49 |
| math-inc/RiemannHypothesisCurves — the Riemann Hypothesis for curves (function-field analogue) | v4.26.0-rc2 | 49 |
| Verified-zkEVM/zkLean | v4.25.2 | 49 |
| math-inc/ZkLinalg — formal security proof of Reed-Solomon IOP-of-proximity constructions | v4.24.0 | 48 |
| CBirkbeck/power_residue_symbols | v4.7.0-rc2 | 44 |
| PnVDiscord/software-foundations-lean — "Software Foundations" ported to Lean 4 | v4.21.0 | 37 |
| jsm28/IMOShortlist2024Lean — 2024 IMO Shortlist formalizations | v4.22.0-rc3 | 34 |
| CBirkbeck/chebotarev-density — the Chebotarev density theorem | v4.32.0-rc1 | 32 |
| stepchowfun/proofs — general formally verified mathematics | v4.33.1 | 26 |
| teorth/newton | v4.5.0-rc1 | 25 |
| teorth/equational — an earlier/smaller companion to equational_theories | v4.12.0-rc1 | 24 |
| Verified-zkEVM/ExtTreeMapLemmas | v4.29.1 | 19 |
| ImperialCollegeLondon/IUM — "Introduction to University Mathematics" course | v4.17.0 | 17 |
| MoonshotAI/CombiBench — combinatorics competition benchmark | v4.24.0 | 14 |
| fpvandoorn/HausdorffSchoolLean — Sept 2023 Hausdorff School tutorial materials, Bonn | v4.0.0 | 14 |
| optpku/CAM-Bench — competition/applied-math proof targets | v4.28.0 | 11 |
| T-Brick/lean-wasm — formalization of the WebAssembly spec | v4.25.0 | 11 |
| frenzymath/reap | v4.28.0-rc1 | 11 |
| frenzymath/jixia | v4.29.0 | 10 |
| YaelDillies/chandra-furst-lipton — corner-free sets / communication complexity | v4.34.0-rc2 | 10 |
| math-inc/KakeyaFiniteFields — the Kakeya set problem over finite fields | v4.26.0-rc2 | 9 |
| WuProver/GroebnerTactic | v4.29.0-rc8 | 8 |
| shetzl/autth — finite automata and context-free grammars | v4.12.0-rc1 | 3 |
| frenzymath/interactive | v4.16.0 | 1 |
tools/harvest.py clones a Lean repository and extracts every theorem and lemma — name, statement and full proof — by a syntactic scan, no build needed; rerunnable against any library at any time.tools/harvest_prove2me.py pulls proved theorems and their accepted solutions from Prove2Me’s API; tools/split_jsonl.py shards any file too big for GitHub.scripts/seed.py, scripts/generate.py, scripts/promote.py: the re-runnable seed, the module generator, and promotion (build the record’s module, then move it to trusted).Tengoku and the wider CompeteMath ecosystem were founded by one person. There is no intention of making money from this project, and it is not affiliated with, nor does it support, any organization, company or political group. No donation or partnership buys influence over the repository’s governance, accessibility, or the integrity of its contents.