AI-assisted inverse theorem for short zero sums in C₃³

For a length-24 positional sequence of nonzero elements of C₃³, every pair of nonempty zero-sum subsequences of length at most three has zero intersection sum exactly when its positions can be permuted into three copies of eight distinct elements with no short nonempty zero sum.

Sources for this record

What was already known

Gao and coauthors proved the threshold η^N(C₃³)=25 and gave extremal constructions, without classifying every length-24 extremal in this form.

AI involvement AI-centered

John Fairfax-Ball directed the research and retains authorship responsibility. His disclosure credits AI with substantive lemma design and the Lean proof; the exact origin of each lemma is not apportioned. No independent human review of this classification is documented.

The project's AI disclosure explicitly covers mathematical research and lemma design for the retained classification, under a human-directed programme; exact per-lemma attribution remains a gap.

Importance Notable

Classifies the extremal sequences one step below the known short-zero-sum threshold.

Limitations

The theorem is restricted to C₃³ and positional length 24. Its broader priority is supported only by the project’s bounded prior-art search.

Technical verification Formal (Lean 4/Mathlib kernel, NanoDa and con-ron)

The registered positional iff is reported checked by Lean and two independent kernels at the cited source. The formal check establishes the encoded claim, not historical novelty.

Formal checking

Statement comparison, proof export and independent kernel replay.

Checker
Palomar Comparator, Lean kernel, NanoDa and con-ron
Scope
InverseZeroSum.Candidate3.frozenClassification, the length-24 positional iff
Basis
Reported by the linked checker
Evidence source

Checking a formal statement does not establish that it matches the original problem.

Human understanding Few

A human-readable direct proof is archived, but no named human account of the central packing/residual mechanism or human proof check is documented; Few is the evidence-gap default.

Correction and publication history