While this submission is a draft, it cannot be used by other submissions.

Concept axioms: adversarial examples and dependency tracking

lax-755887·created ·GitHub @a3f5c47·Lean v4.33.0 epoch · mathlib db584cd6d46c

Loading review…

Sign in with ORCID

Community review

Flags

Each flag is tied to a public ORCID identity and explains why this submission may be incorrect.

No flags have been submitted.

    Community review

    Flag this submission

    State precisely what appears incorrect. This explanation will be public under your ORCID name.

    Abstract

    Unchecked interface claims can yield false application theorems while every proof term remains kernel-correct relative to its assumptions. This submission tests that distinction using persistent queue costs, reused encryption keys, increasing limits of deciders, stuttering simulations, symmetric leader election, and an inconsistent merge-cost bound, adapted from AxiomsAreNonsense.

    The false contracts and their false conclusions are deliberate open obligations. Their derivations should appear in the proof network without establishing the conclusions. Independent controls are proved from the background axioms alone. Additional circular derivations test that mutual support and self-reference do not establish a statement.

    The leader-election example is expressed as a propositional existence claim here. Its original data-valued axiom, its dependent symmetry claim, and the separate imported-axiom reporting example are exercised as validation probes outside the accepted concept package. Those probes distinguish a failure of Lean's diagnostic reporting from acceptance by Lax's independent inspector.

    In the local experiment, all six false application claims and the circular claims remain open, while the independent controls are grounded. The imported-dependency probes are rejected even where Lean's own diagnostic omits the axiom. These examples therefore do not demonstrate an unsound promotion of a conditional proof to a proven statement.

    Concepts

    Concept map
    24 concepts
    100%
    Proven claimOpen claimDefinitionThis submissionA → B: B builds on A

    Proofs

    Proof networkview on GitHub

    100%
    assumptions conclusionProven claimOpen claimStatement 1, 2, … of a claim with several statementsClaim from this submissionProof — open large view for detailsCycle — claims proving each other
    Proof list

    Lean sources for these proofs: proofs/ on GitHub

    Proof code is not displayed; the archive records each proof's checked relationship between claims.

    Related submissions

    No other submission in the archive builds on this one, and this one builds on none.

    Cite this

    This is only the formalizers. The authors of the formalized results may be different (see References).

    @misc{lax-755887,
      title = {Concept axioms: adversarial examples and dependency tracking},
      year = {2026},
      howpublished = {Lax Archive, lax-755887},
      url = {https://laxarchive.org/lax-755887/},
      note = {draft},
    }

    References

    1. Shreyas4991. AxiomsAreNonsense. 2026. Examples inspected at commit 5b95b8b23ea8215f2011469054071fee48812e72. github.com/Shreyas4991/AxiomsAreNonsense

    Discussion

    Ask a question or add context. Endorsements and structured flags are kept in the review panel above.

    Loading discussion…