Skip to content

Erdős Problem #846

Editorial depth: Reviewed explanatory article — evidence qualified.
Snapshot classification: disproved (not automatically the live result).
Subject areas: geometry.

Read this status correctly

The status above is reproduced from GCL's immutable external-catalogue snapshot. It is not a new claim of current mathematical resolution, a source-semantic review, an independent proof check, or GCL certification. A published proof and a GCL replay are separate events.

The question in plain language

The local condition that every finite sample of a planar set has a large no-three-collinear subset need not make the entire infinite set a finite union of such configurations. This is an infinite compactness-style obstruction, not a failure visible in four sample points.

Let A be an INFINITE planar point set such that some fixed epsilon>0 guarantees every finite n-point subset B of A contains at least epsilon n points with no three collinear. Does A decompose into finitely many no-three-collinear classes? NO. The local fractional independence property is not the same as finite chromatic number of the 3-collinearity hypergraph.

A small example

Finite four-point set {(0,0),(0,1),(1,0),(1,3)} has no three collinear; it is trivially one-colorable as a finite geometry instance. Any fixed finite example necessarily has finite partition number, so it cannot witness the infinite counterexample. The finite predicate is nevertheless testable by determinant.

A finite instance tests the definition, not a claim quantified over every large graph, infinite set or arbitrary coloring.

What is known

Maintained https://www.erdosproblems.com/latex/846 read 2026-10-14: source attributes independent 2026 counterexamples to DeepMind and OpenAI internal model, published as Putterman–Sawhney–Valiant [PSV26], On infinite sets with no 3 on a line, arXiv:2602.21275 (2026). It also says Rödl observes a counterexample from Reiher–Rödl–Sales [RRS24] Theorem 1.7 via suitable projection. AI-assisted discovery attribution does not replace proof; full construction/projection not independently source-replayed.

For a finite B in the infinite set A, let alpha(B) be the maximum size of a subset with no collinear triple. The condition is alpha(B)>=epsilon|B| for every finite B with one epsilon. A finite cover by no-three-collinear classes means the collinearity 3-uniform hypergraph of A is finitely colorable. The cited constructions separate positive local fractional independence from finite global chromatic number. Finite configurations alone cannot show such an infinite failure.

GCL's current state

The authenticated original RESULT/1 source return records SOURCE_LOCK_READY and preserves its issue-bound mathematical source observations. This explanatory article is REVIEWED under the exact-hash EC-BULK-005-R6 editorial transaction, subject to the protected merge of PR #2676. Its original source-audit status is retained. Editorial acceptance and public MathJax validation do not certify any original research theorem, external Lean file, or complete historical proof.

Unclosed mathematical proof/source qualifications: PSV26_AND_RRS24_INFINITE_CONSTRUCTION_PROOFS_NOT_REPLAYED; PROJECTION_DETAILS_PENDING.

What remains

Implement exact no-three-collinear test and simple finite subset fraction, then independently inspect 2026 construction and 2024 projection step. Do not claim finite fixtures certify infinite unbounded chromatic number.

For mathematical proof certification, the complete primary argument and any external formalization must undergo separate independent replay. The bounded finite calculation is recorded in the governed R6 fixture; editorial approval does not discharge the outstanding research obligations.

Sources and claim boundaries

  • Erdős problem #846 and maintained LaTeX/bibliography, read 14 October 2026.
  • Authenticated source-audit return, including original historical bibliography and source proof limitations.
  • Pinned teorth/erdosproblems revision 536408716724a2402bdfa7c1d764bf83d64a8c89, data/problems.yaml. External proofs not independently GCL certified.
  • Editorial decision: exact-hash reviewed explanatory article, protected R6 transaction. Mathematical theorem proof / external Lean certification: NOT CLAIMED. The original SOURCE_LOCK_READY return, primary-source attribution, and proof-review holds remain.

Catalogue and evidence provenance

  • Pinned catalogue source: teorth/erdosproblems at 536408716724.
  • GCL issue: #2117 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
  • GCL triage: OUTSIDE_STRICT_OPEN_TRIAGE.
  • Formal-statement coverage: A provider statement was included in the pinned snapshot. Presence does not mean the statement or proof has been checked.
  • External mathematical status: consult current problem page and its dated bibliography.

Back to catalogue · Compendium introduction