This repository supports the Zenodo preprint:
Stress Testing Frontier AI Systems for Verifiable Reasoning and Code Generation: GPT-5.5 Pro / Extra High, Codex, Claude Opus 4.7 Chat, and Claude Code
Author: Marcel Jiron / MarMarLabs
Zenodo DOI: https://doi.org/10.5281/zenodo.19779656
Artifact repository: https://github.com/MarMarLabs/frontier-ai-stress-test-audit
This repository preserves the source files, audit data, local test logs, and verification notes for a reproducibility-oriented audit of frontier AI systems.
The paper is not a controlled laboratory benchmark. It is an artifact-based audit of submitted outputs, source-code artifacts, transcripts, and test logs. Claims should be read together with the available artifacts and verification notes.
paper/: preprint PDF, LaTeX source, and bibliography.data/: summarized evaluation tables used in the paper.verification/logs/: local test logs and command outputs.verification/scripts/: scripts used for manual or local checks.metadata/: manifest describing the included artifacts.
Jiron, M. (2026). Stress Testing Frontier AI Systems for Verifiable Reasoning and Code Generation: GPT-5.5 Pro / Extra High, Codex, Claude Opus 4.7 Chat, and Claude Code. Zenodo. https://doi.org/10.5281/zenodo.19779656
Unless otherwise stated, text, tables, and non-code research artifacts are released under CC BY 4.0.
Code files, if any, should be checked for their own license headers.
Third-party content remains under its original license.