CFC+HAWM — AI Evaluation and Regression Pack v1.1

CFC+HAWM AI Evaluation and Regression Pack v1.1 is a self-contained package for independent evaluation of selected closure and evidence-state reasoning cases in large language model outputs. The package includes a 20-case provider-neutral regression suite, a Windows regression runner, final Gemini API diagnostic results, a Claude manual diagnostic summary, two highlighted challenge cases (R010 and R015), evaluator guidance, step-by-step instructions, and SHA256 checksums. In the automated Gemini API regression run, 18 of 20 cases produced stable expected labels, one case showed closure-label instability (R010), and one case produced a reproducible semantic failure (R015). These results are diagnostic and should not be interpreted as a general model accuracy rate. The package clearly separates ordinary model output (MODEL_REPLY_UNCHECKED) from frozen CFC authorization. Frozen CFC was not modified. A simple Windows launcher and instructions for creating a Gemini API key are included to make independent reproduction easier for non-technical users and AI evaluators.