SEASON 2 — Paper Artifacts and Reproducible Outputs (TVL1 vs ORTHO, MP/NMP Gating)

This record contains the complete Season-2 dataset and minimal code to reproduce all paper figures and tables for the comparative study of non-orthogonal TV+L1 and orthogonal decomposition models under identical scoring. Amplitude |H| is preserved; phase is decomposed into minimum-phase (MP) and all-pass residual (NMP/Field-2). We provide mixed-reference FRFs, per-model outputs, validation summaries (KK error, sMAPE, coverage, Q-consistency, −3 dB censoring), and publication-ready figures with their source CSV. A one-click Colab/Linux pipeline rebuilds all artifacts in minutes—no heavy simulations required. What’s inside Inputs: inputs/FRF.csv, optional inputs/windows/state_windows{1..5}.csv One-click pipeline (scripts): 1.PREP_Unzip.py … 9.PAPER_ARTIFACTS.py Mixed reference: outputs/se2_mix/* (RAW/MP tracks, mix_reference_report.json) Model outputs: outputs/core/ (TVL1 / ORTHO one-cell results) Validation & summaries: outputs/reports/summary_VALIDATION.json, summary_MINIMAL.*.json, summary_STACK.json Paper assets: outputs/paper/*.png, outputs/paper/*.csv, outputs/paper/paper_summary.json Provenance: outputs/MANIFEST.sha256 (SHA-256 checksums), outputs/ENV.json (runtime) Reproducibility (3 steps) Run the one-cell pipeline (scripts 1..9) on Colab/Linux to regenerate core outputs. Validate: ensure outputs/reports/summary_VALIDATION.json shows "pass": true (KK/sMAPE/coverage/Q/−3 dB). Package: confirm outputs/MANIFEST.sha256 and paper figures/CSVs under outputs/paper/. Winner rule (brief) Winner is decided by MP/NMP phase gating (KK/NMP thresholds) and sMAPE; ties break by lower sMAPE. Phase residuals (all-pass) quantify Field-2.