Proof of Pattern: Diagnostic Observability and Documented Proxy-Level Structural Effects under Controlled PLA-Derived Constraints
Proof of Pattern is a prompt-based proxy study examining whether selected PLA-derived constraints make coherence-relevant structural relations more observable and action-binding in existing large language models under controlled contradictory conditions. Across two multi-round crisis scenarios and heterogeneous model families, the study compares genuine conditions without PLA initialization with PLA-derived observational conditions. The analysis focuses on mismatch preservation, temporal drift, cross-scale consistency, restricted commitment, and Regulated Slowdown, while distinguishing explicit structural representation from functional realization across evolving trajectories. The findings document both recurrent proxy-level structural effects and their boundaries. PLA-derived conditions increased the explicit observability of mismatch, drift, cross-scale relations, and constraints on irreversible commitment. In selected complete trajectories, these relations also became action-binding. At the same time, the corpus documents cases in which explicit structure ceased to govern action, slowdown became persistent non-response, physical feasibility was violated despite being represented, action history was retrospectively replaced, or evaluator scores failed to distinguish formal markers from functional realization. The study therefore establishes diagnostic observability together with documented proxy-level structural effects and mapped failure boundaries. It does not claim implementation or validation of PLA as an internal model architecture, a final architecture, or an established safety mechanism. The empirical basis comprises 3,699 archived evidence files documenting 48 runs and 69 evaluations. Primary runs, intra-model and cross-model evaluations, repetitions, initialization records, JSON and Markdown exports, PDFs, editable documents, screenshots, and associated study artifacts are preserved in the linked Evidence Archive. Associated resources Evidence ArchiveDOI: 10.5281/zenodo.23125191 Precision-Layer Architecture Working PaperDOI: 10.5281/zenodo.18642675 Foundation PaperDOI: 10.5281/zenodo.23124273
Authors
- Diana Becker
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-10-05
- DOI
- https://doi.org/10.5281/zenodo.23124964
- Primary Topic
- Psychiatry, Mental Health, Neuroscience
- Type
- article
- Field-Weighted Citation Impact
- 0.00