Empirical, comparative evaluation records for Symbolic Mechanics / Human OS under controlled adversarial conditions across multiple large language model platforms.
The benchmark line is an evaluation surface. It compares interpretive performance under fixed constraints and preserved corpora. It does not function as clinical validation, therapeutic outcome study, or commercial benchmark claim.
Benchmark records sit beside the specification, theory-publication, and prototype lines as a distinct public evidence route.
Five-platform equal-weight aggregate across ChatGPT, DeepSeek, Claude, Grok, and Gemini, using a preserved 24-item corpus.
Core result: B won 23/24 vote-based cases; overall aggregate mean B 48.13 vs A 45.07.
Formal aggregate report for the completed Phase I benchmark line.
Specification Line · Theory Publications · Prototype Records