Semantic Austerity Testing (SAT): A Canonical Methods Reference for Interaction-Protocol Robustness in Large Language Models

This record provides the canonical methods reference for Semantic Austerity Testing (SAT), a qualitative robustness framework for evaluating large language models (LLMs) under conditions of intentionally reduced semantic signal, denied instrumental purpose, and suppressed engagement incentives. SAT evaluates restraint rather than performance, focusing on a model’s ability to tolerate silence, ambiguity, absurdity, and non-obligation without fabricating intent, authority, or conversational momentum. The methodology formalized here underpins a broader SAT research corpus examining monotony tolerance, semantic null handling, recursive instruction overload, prompt-less activation, humor blindness, and authority hallucination in both local and hosted LLMs. This release is intended to serve as a stable, reusable methodological citation, allowing future SAT studies to omit repeated methods exposition while maintaining continuity across publications.