google-deepmind / google-deepmind/formal-conjectures
Mark eqSystem6_no_solution_d3 as solved via external Lean 4 proof
- Dominant language
- Lean
- Stars
- 1.3k
- Forks
- 485
- Avg merge
- 1d 20h
- Merged PRs (30d)
- 328
Description
A complete external Lean 4 proof is now available for MonochromaticQuantumGraph.eqSystem6_no_solution_d3.
Pinned proof:
https://github.com/algal/krenn-gu-6x3-certificate/blob/c04696e515e0c02be140353fb52ea60c62e827b1/KrennGuCertificate/Unrestricted.lean#L201-L222
Repository and verification instructions:
https://github.com/algal/krenn-gu-6x3-certificate/tree/c04696e515e0c02be140353fb52ea60c62e827b1
The proved proposition is exactly the right-hand side of the existing benchmark theorem:
¬ ∃ W : MonochromaticQuantumGraph.WeightsN 6 3 ℂ,
MonochromaticQuantumGraph.EqSystemN 6 3 W
A clean clone of the pinned release passed all 50 committed artifact checksums and all 8,421 Lean build jobs. The final axiom report is:
[propext, Classical.choice, Lean.ofReduceBool, Lean.trustCompiler, Quot.sound]
There is no sorryAx or custom axiom in the proof. The Lean.ofReduceBool and Lean.trustCompiler dependencies arise from native_decide and are documented in the repository trust boundary.
Following CONTRIBUTING.md, I plan to submit a small status PR that:
- changes category research open to category research solved;
- replaces answer(sorry) with answer(True);
- adds a formal_proof using lean4 attribute pointing to the pinned proof above.
The long proof remains in its external repository, as prescribed by the formal_proof mechanism.
AI-use disclosure: the proof and this registration contribution were developed collaboratively with OpenAI Codex. Its correctness claim rests on the committed Lean verification and reproducible certificate checks, not on model-generated prose.
Contributor guide
Research direction
Read CONTRIBUTING.md and inspect the existing benchmark theorem and formal_proof conventions. Verify the pinned KrennGuCertificate/Unrestricted.lean proof using the repository instructions, then make the status, answer, and formal_proof metadata changes; done means the checks and Lean build pass without introducing a sorryAx or custom axiom.
Written by the indexing model from the issue text.
Assessment
- Domain
- tooling
- Issue type
- Feature
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Activity status
- Quiet
- Clarity
- Clearly specified
- Newbie friendliness
- 68/100