Published June 12, 2026 | Version v0.2 Pre-Print

Regime Convergence in Self-Revising Discovery Systems

Description

This preprint gives a finite obstruction-calculus response to categorical models of self-revising discovery systems. The central claim is that categorical regime transition can identify residual content, but residual detection is not yet certification. Under declared finite-dimensional rational carriers, exact realization, replay, bounded budget, positive operation cost, and a total terminalizer, transported residuals can be routed to explicit terminal statuses: ExactRepair, StabilizedSurvivor, OpenObstruction, LicensedLift, BudgetUnknown, NoRule, or Refused.

The paper positions Wang–Buehler-style categorical transport as a useful grammar for candidate discovery episodes, then supplies the missing finite residual-terminalization boundary. It distinguishes transport residuals from authority-bearing terminal status, separates MDL/AIC or benchmark gates from certificates, and gives a finite realization-compatibility lemma connecting transported cokernel residuals to FOC quotient residuals under exact image-matching hypotheses.

The release is supported by a Paper022 Lean4 stack of 37 files covering quotient-bridge kernels, terminalizer routing, gate/certificate separation, numeric-gate provenance, discovery certificate packets, authority firewalls, production-manifest interfaces, audit gates, and negative examples. These mechanizations certify local finite interfaces and examples; they do not claim production-scale scientific discovery, empirical validation, global convergence, or physical/quantum authority.

Files

022_RE~1.PDF

Files (986.8 kB)

Name Size Download all
md5:9ebb7462595e6a530d85bd196552b1e0
463.6 kB Preview Download
md5:d79aa6809490bbaa440cf32f5d735635
48.6 kB Preview Download
md5:bd27e135fbfc7da6a7beb0bfe9a9fa3a
474.5 kB Preview Download

Additional details

Additional titles

Subtitle
A Finite Obstruction-Calculus Formulation

References

  • Wang, F. Y., & Buehler, M. J. (2026). Self-Revising Discovery Systems for Science: A Categorical Framework for Agentic Artificial Intelligence. arXiv preprint arXiv:2606.01444