Gauntlet arXiv submission preparation

Upload cortex-gauntlet-arxiv.zip, not the rendered PDF. Select main.tex and pdfLaTeX.
arXiv currently defaults to TeX Live 2025. Inspect its generated PDF before submitting.
The source archive contains main.tex, the official ICLR 2027 style files and 8 PDF figures.
The bibliography is embedded; the document is a named preprint, not an ICLR submission.
No external conversion, Python, network, shell escape or private data is required.
Confirm author details and coauthor consent; choose an appropriate category and license.
Account registration, possible endorsement and moderation remain arXiv's requirements.
No arXiv identifier or acceptance is claimed.

Title: Cortex: A Fixed-Point Theory of Governed Coding Agents
Authors: Marius-Constantin Dinu, Florian Zeba (Alpha Omega Labs)
Abstract: A coding agent combines a large language model (LLM) with a harness that plans, edits, and executes code. We study Cortex, a supervisory layer for pre-execution governance, requirement tracking, and iterative validation and repair. We formalize requirement closure on a finite lattice. For a fixed requirement set and a sound, monotone validation rule with persistent evidence, exhaustive iteration reaches the least fixed point with no more strict increases than there are requirements. Fair consequence scheduling reaches the same closure. A separate stochastic model bounds expected completion time under a uniform positive-progress assumption. These results depend on their stated assumptions; artifact repair, fallible validation, and finite execution budgets do not satisfy them automatically.

We define Gauntlet's trajectory-similarity and build-gated scores, attack-success estimators, Wilson intervals, and case-cluster bootstrap summaries. Effective-feedback accounting counts newly certified requirements, rather than estimating their information value or utility. The five evaluation families support raw-versus-governed comparisons. We report retained benchmark observations, including rescoring of existing artifacts; constructed outcomes are excluded. Unequal observation sets and incomplete provenance prevent causal claims about governance or measured long-horizon scaling.
Comments: 27 pages, 8 figures, 9 tables. Paper and latest results: https://benchmark.cortex.a2olabs.com . Benchmark source: https://github.com/Xpitfire/cortex-gauntlet .

Suggested primary category for author review: cs.AI. Leave journal reference and DOI blank unless assigned.
Recompile after extracting the ZIP: pdflatex -no-shell-escape main.tex (run twice).

Official instructions:
https://info.arxiv.org/help/submit/index.html
https://info.arxiv.org/help/submit_tex.html
https://info.arxiv.org/help/faq/texlive.html
https://info.arxiv.org/help/prep.html