google-deepmind / google-deepmind/formal-conjectures
Steiner system S(6,7,23): four automorphism groups eliminated via Kramer-Mesner
- Dominant language
- Lean
- Stars
- 1.3k
- Forks
- 485
- Avg merge
- 1d 20h
- Merged PRs (30d)
- 327
Description
Partial progress on Open Subproblem 1 from Epoch AI's FrontierMath
(https://epoch.ai/frontiermath/open-problems/large-steiner-systems):
construct a Steiner system with t > 5.
Using Kramer-Mesner reductions, I have computationally eliminated four
candidate automorphism groups for S(6,7,23):
- M₂₃: trivial 3×4 KM matrix, no exact cover (category: research solved)
- AGL(1,23): structural uncoverable row (category: research solved)
- Z₂₃⋊Z₁₁: exhaustive DLX search, 11M nodes, UNSAT (category: research solved)
- D₂₃: divisibility-2 obstruction (category: research solved)
The Z₂₃ case (4389×10648 KM matrix) remains open and is currently
running under CP-SAT with 80 workers (category: research open).
Related to #3524. References the FrontierMath large Steiner systems
open problem. I plan to contribute Lean formalizations of these results.
Contributor guide
Research direction
The issue names no repository files, tests, or Lean entry points. Start by reading the linked FrontierMath problem and related issue #3524, then review the four Kramer–Mesner eliminations and the open Z₂₃ case. Done means formalizing the reported results in Lean, as proposed in the issue.
Written by the indexing model from the issue text.
Assessment
- Domain
- tooling
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100