Formal verification framework validates mathematical proofs of self-improving systems in artificial intelligence theory, ensuring complete reproducibility via automated audit controls.
Material supporting the paper An Axiomatic Geometry of Self-Improvement: Phase Structure, Sustainment Ceilings, and the Dimensional Failure of Post-Hoc Alignment. Three components. A Lean 4 development of the axiom system and the principal results, built against a pinned Mathlib revision, free of incomplete proofs, with the axiom footprint and the elaborated signature of each load-bearing result pinned by guarded assertions. Python scripts reproducing every numerical quantity the paper reports. And the audit apparatus: checks that every declaration the paper names resolves, that every cited result reaches the axiom system or is recorded as deliberately abstract, that the recomputed equations agree, and a suite of negative controls that breaks each guarded property on purpose and confirms the corresponding check fails for the right reason. Licensing is split: the paper and figures under CC BY 4.0, the Lean development, the analysis scripts and the audit tools under Apache 2.0.
No takes yet. Share an insight, caveat, or question.
Yong Yang (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: