Dataset for "Computational Validation of Bounded Coherence: From Biological Oscillators to AI Content Generation" (11,394 Generation–Verification Cycles)

This dataset supports the preprint: “Computational Validation of Bounded Coherence: From Biological Oscillators to AI Content Generation” (Phillips, 2026). It contains structured telemetry from 11,394 successful generation–verification cycles collected across six AI content modalities: Text Image Audio Code Video Music The dataset operationalizes computational phase coherence using the balance metric: S(W)=tverificationtgeneration+tverificationS(W) = \frac{t_{verification}}{t_{generation} + t_{verification}}S(W)=tgeneration+tverificationtverification Each record corresponds to a completed generation–verification loop and includes balance, stability metrics, modality labels, and regime classification used for Beta distribution fitting and chaos-boundary analysis. Files Included raw_data.json — Full record-level dataset (11,394 records) per_modality_stats.json — Summary statistics per modality (mean, σ, cv, α, β) cross_modal_analysis.json — Aggregated cross-modal comparisons paper3_summary.md — Human-readable summary of dataset structure MANIFEST.txt — SHA256 checksums for integrity verification README.md — Documentation and field definitions Total dataset size: ~12.5 MB Purpose This dataset enables independent replication of: Beta-distributed balance fitting across modalities Modality-specific α parameter estimation β–cv inverse relationship analysis Identification of a computational chaos boundary in code generation (cv ≈ 0.92, α < 1) The dataset tests whether bounded phase coherence, previously derived analytically in biological oscillator systems, generalizes to computational substrates. Related Works Analytical foundation: 10.5281/zenodo.18235945 10.5281/zenodo.18616949 Associated preprint: [Insert Paper 3 DOI after publication] Reproducibility All files are provided in open JSON and Markdown formats.Checksums are included for verification. No personally identifiable information is contained in this dataset.