Erdős Problem #571: rational exponents for bipartite Turán numbers
Erdős problem #571 · erdosproblems.com/571
For every rational , there exists a finite bipartite graph such thatEquivalently, every rational exponent between and occurs as the order of growth of the Turán number of a single bipartite graph.
- Result
- Proved(see note)
- Status
- Resolved
- AI contribution
- AI-discovered
- Method
- Argument
- Field
- Graph theory
- Posed by
- Paul Erdős, Miklós Simonovits
- Year posed
- 1974
- Years open
- 52y
- Solved
- 2026-08-26
- Model
- GPT-6 Astra (pre-release)
- Vendor
- OpenAI
- Collaborators
- Tom Adamczewski
- Verification
- Lean-verified
- Publication
- Announced
- Significance
- 52 / 100
- Disclosed cost
- $617
- Wikipedia
- No dedicated article
What was actually shown
For every rational satisfyingthe Lean theorem constructs some finite and a bipartite graphsuch thatas .
This resolves the single-graph rational-exponents conjecture. Earlier work of Bukh and Conlon proved the corresponding statement only for a finite family of forbidden bipartite graphs, and subsequent work realized many large classes of individual rational exponents. Astra's theorem covers every rational with one forbidden graph for each exponent.
What the AI did
A pre-release GPT-6 Astra autonomously found the proof and wrote its Lean formalization in the FrontierMath Erdős benchmark, with no human seeing or steering the proof search. The roughly 10,000-line development constructs balanced rooted graph models for every rational exponent and proves closure operations that preserve the required extremal-number behavior, including edge subdivision by paths of arbitrary length, adding hubs to the two colour classes, and positive rooted powers. These constructions are combined to realize every rational by a single finite bipartite graph.
Verification
Lean-verified. Checked here on 6 September 2026 from a clone of tadamcz/erdos571 at 661cc1d: 10,460 lines of Lean, zero sorry outside the statement stubs, zero axiom declarations, no native_decide, unsafe or implemented_by; Comparator configuration present and CI runs it with only propext, Quot.sound and Classical.choice. The statement was autoformalized for the benchmark and reviewed by Thomas Bloom; it uses mathlib's extremalNumber and IsBipartite and Asymptotics.IsTheta, and the repository's README compares it to the informal statement. One grep hit for the word 'externally' in a docstring is the only match for the risky-feature scan. erdosproblems.com, the field's own record, marks the problem PROVED (LEAN) with a proof exposition by Thomas Bloom, which is why this is Resolved rather than Candidate: the canonical tracker has accepted it.
Sources
- Lean proofSolution.lean and the Lean development
- Lean statementChallenge.lean: the compared statement
- CodeGithub
- AnnouncementEpoch AI, Announcing FrontierMath Erdős (1 September 2026)
- Problem recorderdosproblems.com/571: status and Thomas Bloom's proof exposition
Submitted by VibeGene on