Reproducibility · manifests
Reproduce this lab
Every experiment, simulation, derivation, training run, scan, or analysis ships a manifest with sealed inputs (external data with links, or repo-pinned internal artifacts), exact entrypoints, compute venue, an estimated cost and wall-clock, and a verification method — so a stranger, or Hubify, can reproduce it without asking us anything. The full-reproduction pass across every program is the final pre-publication test of this lab, not an afterthought.
3Research programs
112Experiment manifests
100Runnable now
$36.04Est. total reproduction cost
10 experiments need a data restore before they can run; 2 are superseded and kept for lineage only, never offered as a live reproduction target. Per-program cost/wall-clock totals are rollups, not literal sums of the experiment rows below — see each program's full_reproduction.order note for sequencing.
Research program · anomaly-discovery
DESI anomaly discovery
Question: What unusual spectra emerge from a full-scale DESI anomaly search, and which candidates survive scientific validation?
29 experiments · 23runnable now · full reproduction $7.74 · The completed AUG-011 scan ran about 45.5 hours on a RunPod A4000 and verified 36,634 shard receipts. Full independent reruns remain download-bound and multi-day; downstream sample selection, validation, taxonomy, and manuscript work are still pending because the full shard corpus is not in this checkout.
Experiments — reproduction order
| Experiment | Inputs | Scripts | Venue | Est. cost | Wall-clock | State |
|---|---|---|---|---|---|---|
BigAE enhanced 18M/22.5M-row DESI inference (historical, unreconciled) anomaly-bigae-18m-inference-historical | best_model_47k.pt (archived BigAE checkpoint, 496->512->256->128) | python3 enhanced_18M_inference.py | not recommended for reproduction — superseded by the clean_rerun campaign (see anomaly-clean-rerun-scan) | free (local) | n/a | needs data restore |
Silver crossmatch (2,145-row SNR-filtered slice) anomaly-silver-crossmatch · depends on anomaly-bigae-18m-inference-historical | internal derivation — no external input | python3 silver_crossmatch.py | local | free (local) | minutes to low hours (depends on external crossmatch service throughput) | runnable now |
Uncataloged taxonomy (1,127 objects, 10 families) anomaly-uncataloged-taxonomy · depends on anomaly-silver-crossmatch | internal derivation — no external input | python3 classify_uncataloged.py | local | free (local) | minutes | runnable now |
Injection recovery test (per-class completeness / false-positive rates) anomaly-injection-recovery-test · depends on anomaly-bigae-18m-inference-historical | internal derivation — no external input | python3 injection_recovery_test.py --n-per-class 200 --seed 42 | local | free (local) | minutes | runnable now |
NEOWISE crossmatch (IR variability, 16/283 meet variability rule) anomaly-neowise-crossmatch · depends on anomaly-silver-crossmatch | internal derivation — no external input | python3 neowise_crossmatch.py python3 neowise_crossmatch_silver.py | local | free (local) | minutes to low hours (depends on external IRSA/NEOWISE query throughput) | runnable now |
Gold anomalies — z6 QSO spectra (12 DESI Redrock z>6 QSO candidates) anomaly-gold-z6-qso-spectra · depends on anomaly-uncataloged-taxonomy | internal derivation — no external input | python3 download_z6_qso_spectra.py python3 replot_z6_spectra.py | local | free (local) | minutes to low hours (DESI spectra download-bound) | runnable now |
Photo-z from latent vectors (supervised MLP, sigma_NMAD=0.0279) anomaly-photoz-latent-vectors · depends on anomaly-bigae-18m-inference-historical | 22.5M-row enhanced DESI catalog latent features (BigAE encoder output) | python3 train_photo_z.py | local (blocked pending data restore) | free (local) | n/a until parent latent features are restored | needs data restore |
f_NL tracer selection / step4-6 bias validation + alpha empirical calibration (negative result) anomaly-fnl-tracer-selection | internal derivation — no external input | python3 step4_bias_validation.py python3 wave_14_vvv_alpha_empirical.py | local | free (local) | minutes | runnable now |
clean_rerun campaign (AUG-011) — completed sealed generation, 2026-08-07 anomaly-clean-rerun-scan | best_model_47k.pt (archived BigAE checkpoint, 496->512->256->128, reused input) | python3 derive_locator_inventory.py derive --zcatalog zall-pix-iron.fits --output locator_inventory_draft.jsonl python3 build_calibration.py --zcatalog zall-pix-iron.fits --manifest run-contract.json --model best_model_47k.pt --coadd-cache-dir <cache> --training-manifest-output training_manifest.json --validation-manifest-output validation_manifest.json --output calibration.json python3 run_scan.py --inventory locator_inventory.jsonl --contract run-contract.json --model best_model_47k.pt --shard-dir <shards> --receipt-dir <receipts> --checkpoint <ckpt> --coadd-cache-dir <cache> python3 clean_rerun_contract.py build --model best_model_47k.pt --inference-code enhanced_18M_inference.py --input-manifest input_manifest.json --calibration calibration.json | runpod (A4000-class GPU or CPU-strong instance, ~200GB volume) | $7.74 | Completed run recorded at 45.5h on RunPod A4000 tc291bka0r6fl3; no separate estimate remains for the sealed generation. | runnable now |
DP3-15 held-out re-inference (structural-ceiling demonstration) p3-dp3-15-heldout | r42_phase2 5-seed BigAE ensemble checkpoints (496->128) | python3 dp3_15_heldout_reinference.py | local | free (local) | 30-60 minutes (dominated by the live SPARCL re-pull of ~20,000 held-out spectra) | runnable now |
6-way / 7-way / 8-way positional dedup (275,151 -> 269,317 unique, 2.12% collapse) p3-positional-dedup | python3 reproduce_headline_dedup.py python3 sixway_dedup.py python3 pathc_positional_dedup.py | local | free (local) | minutes | runnable now | |
DESI 5-fold cross-validation reproducibility gate (mean pairwise Jaccard 0.862) p3-kfold-cv-gate | internal derivation — no external input | python3 train_desi_kfold.py --training-shards <shards> --output-dir outputs/desi_kfold --n-folds 5 --seed 20260420 python3 fetch_desi_47k_training.py python3 heldout_tail_preservation.py | local | free (local) | 1-3 hours (5-fold model training, 47,000-row pool) | runnable now |
Planck held-out membership test + native re-inference (partial, 48/200 vs 30 expected) p3-planck-heldout-membership | best_cmb_native.pt (native CMB autoencoder checkpoint) cmb_native_patches.npy (200k masked SMICA patch tensor) | python3 held_out_rescore.py | local (membership test only; full native re-inference blocked) | free (local) | minutes for the membership test; full native re-inference is not schedulable until best_cmb_native.pt + cmb_native_patches.npy + cmb_native_all_scores.parquet are re-staged | needs data restore |
eROSITA scaler-leakage bounded control (top-298 overlap 257/298, J=0.76) p3-erosita-scaler-leakage-control | eROSITA DR1 feature table (930,203 sources, pod-side only) | generating script not identified in current repo tree (grep -l "erosita_scaler_refit" over pipelines/p3_anomaly_engine/**/*.py found no match — the two closest candidates, erosita_membership_reproduce.py and r24conf_erosita_axis_sweep.py, write to differently-named output files: outputs/erosita_membership_reproduce.json and r24conf_erosita_axis_sweep.json respectively, not erosita_scaler_refit.json) — this JSON result file is the surviving artifact and is inspectable/re-analyzable as-is | runpod (pending generating-script recovery) | free (local) | unknown until the generating script is recovered or rewritten from the committed result JSON's documented method | runnable now |
NANOGrav 15-yr free-spectrum PTA MCMC (real Zenodo KDE likelihood, emcee) p3-nanograv-pta-mcmc | BIGBOUNCE_WORKSPACE=/path/to/writable/dir python3 emcee_freespec.py | local | free (local) | under 1 minute (committed run reports production_seconds=24.97 for 32 walkers x 10,000 production + 2,500 burn-in) | runnable now | |
Multi-survey summary / crossmatch / spatial-clustering / score-distributions (8 surveys) p3-multisurvey-summary-crossmatch | internal derivation — no external input | python3 bulk_cross_match_all.py | local (script survives; per-survey raw output regeneration untested since the 2026-04-08 snapshot) | free (local) | hours (8-survey crossmatch, external archive query throughput bound) | needs data restore |
UMAP multi-seed stability (Pod 1 production) p3-umap-multiseed-stability · depends on p3-multisurvey-summary-crossmatch | internal derivation — no external input | no generating script preserved in repo — historical result artifact only; full reproduction requires restoring the lost UMAP pipeline script from the pod1_namaster_umap production run | unknown — blocked until the UMAP pipeline script is restored | free (local) | unknown | needs data restore |
S>8 flagship anomaly sample enrichment (photometry/spectroscopy/morphology join) [SAMPLE-V1, provenance under review] p3-flagship-s8-enrichment | flagship_sample_s8.parquet DESI zcatalog/target coordinates | python3 build_flagship_sample.py --enrich (pod-side; see phase3.log) | runpod-a4000-or-equivalent | $1.70 | ~10h | runnable now |
AllWISE (VizieR) photometry join for the S>8 enriched sample [SAMPLE-V1, provenance under review] p3-flagship-s8-allwise-photometry · depends on p3-flagship-s8-enrichment | flagship_sample_s8_enriched.parquet | python3 build_flagship_sample.py --wise-join (pod-side; see phase3.log) | local-cpu | free (local) | ~2h | runnable now |
SIMBAD/NED positional cross-match of the S>8 enriched sample [SAMPLE-V1, provenance under review] p3-flagship-s8-simbad-ned-crossmatch · depends on p3-flagship-s8-enrichment | flagship_sample_s8_enriched.parquet | python3 build_flagship_sample.py --crossmatch-simbad-ned (pod-side; see phase3.log) | local-cpu-or-runpod | free (local) | ~9-10h (network-bound VizieR/SIMBAD/NED rate limits) | runnable now |
Descriptive taxonomy (UMAP + clustering, Q1 labels) of the S>8 unmatched population [SAMPLE-V1, provenance under review] p3-flagship-s8-taxonomy · depends on p3-flagship-s8-simbad-ned-crossmatch | flagship_crossmatch_unmatched.parquet | python3 build_flagship_sample.py --taxonomy (pod-side; see phase3.log; sklearn 1.9.0, umap_learn 0.5.12, numpy 2.4.6) | local-cpu | free (local) | ~1-5min | runnable now |
S>3 flagship anomaly sample enrichment (photometry/spectroscopy/morphology join) [SAMPLE-V2, science-only, contamination-fixed] p3-flagship-v2-enrichment | flagship_sample_v2.parquet DESI zcatalog/target coordinates | python3 enrich_flagship_sample.py (pod-side, stage 04_ENRICH; see pod_phase3_v2.sh) | runpod-a4000-or-equivalent | $0.50 | ~2h45m | runnable now |
AllWISE (VizieR) photometry join for the S>3 v2 science-only sample p3-flagship-v2-allwise-photometry · depends on p3-flagship-v2-enrichment | flagship_sample_v2_enriched.parquet | python3 wise_join_flagship.py (pod-side, stage 06_WISE; radius_arcsec=3.0, astroquery.vizier.Vizier) | local-cpu | free (local) | ~40-45min | runnable now |
SIMBAD/NED positional cross-match of the S>3 v2 science-only sample p3-flagship-v2-simbad-ned-crossmatch · depends on p3-flagship-v2-enrichment | flagship_sample_v2_enriched.parquet | python3 crossmatch_flagship.py (pod-side, stage 05_CROSSMATCH; radius_arcsec=3.0) | local-cpu-or-runpod | free (local) | ~3-4h (network-bound VizieR/SIMBAD/NED rate limits, 1.0s sleep between queries, radius 3.0 arcsec) | runnable now |
Descriptive taxonomy (UMAP + clustering, 25 clusters -> 8 families) of the S>3 v2 science-only unmatched population p3-flagship-v2-taxonomy · depends on p3-flagship-v2-simbad-ned-crossmatch | flagship_crossmatch_v2_unmatched.parquet | python3 taxonomy_flagship.py (pod-side, stage 07_TAXONOMY; umap_neighbors=15, umap_min_dist=0.05, score_tier_quantiles=[0.5,0.8,0.95]) | local-cpu | free (local) | ~1-5min | runnable now |
Known-object recovery benchmark for the anomaly flagship catalogue (ledger item #8) anomaly-known-object-recovery-benchmark | Sealed locator inventory (HEALPix footprint definition) Reference 'unusual object' classes (Baron & Poznanski 2017, Roma-BZCAT, BALQSO, CV, carbon-star, LAE, EELG, changing-look-QSO, SLSN-host, GRB-host catalogues) | python3 benchmark_known_object_recovery.py --fetch-references --reference-cache-dir <cache> && python3 benchmark_known_object_recovery.py --crossmatch --reference-cache-dir <cache> --catalogs-config <config.json> --locator-inventory sealed_2026-08-05/locator_inventory.jsonl --out-dir <out> | local CPU or the phase-3 RunPod pod (same host, no extra GPU need) | free (local) | minutes for --fetch-references (VizieR query latency dominates, ~10-30s per class with checkpointless single-shot fetch); seconds to low minutes for --crossmatch once the flagship sample carries target_ra/target_dec (via enrich_flagship_sample.py) for the full S>5/S>8 samples | needs data restore |
Known-object recovery benchmark (ledger item #8) run against the S>3 v2 science-only flagship sample anomaly-known-object-recovery-benchmark-v2 · depends on p3-flagship-v2-enrichment | Sealed locator inventory (HEALPix footprint definition) flagship_sample_v2_enriched.parquet (S>3 science-only sample, n=1244) Cached reference 'unusual object' classes (BAL quasars, Roma-BZCAT blazars, CV/WD binaries, LAEs, SLSN hosts) | python3 benchmark_known_object_recovery.py --crossmatch --reference-cache-dir ~/Desktop/CODE_YOU/bigbounce_datasets/aug-011-clean-rerun/recovery_refs_2026-09-02 --reference-manifest ~/Desktop/CODE_YOU/bigbounce_datasets/aug-011-clean-rerun/recovery_refs_2026-09-02/reference_manifest_local.json --catalogs-config /tmp/catalogs_config_v2.json --locator-inventory pipelines/p1_highz_tracers/clean_rerun/sealed_2026-08-05/locator_inventory.jsonl --radius-arcsec 1.5 --out-dir pipelines/p1_highz_tracers/clean_rerun/results_2026-08-07/phase3_v2/recovery_benchmark | local CPU (any host with the cached reference_catalogs, or a fresh --fetch-references run from a host that can reach vizier.cds.unistra.fr's TAP query endpoint) | free (local) | under 1 minute for --crossmatch once references are cached; 10-40 min for a fresh --fetch-references pass | runnable now |
Ledger #8 known-object recovery benchmark (VizieR reference classes vs S>8 sample) [SAMPLE-V1, deferred] p3-ledger8-known-object-recovery-benchmark · depends on p3-flagship-s8-enrichment | flagship_sample_s8_enriched.parquet VizieR reference class cache (5 fetched classes) sealed_2026-08-05/locator_inventory.jsonl (DESI footprint) | python3 benchmark_known_object_recovery.py --crossmatch --reference-manifest ~/Desktop/CODE_YOU/bigbounce_datasets/aug-011-clean-rerun/recovery_refs_2026-09-02/reference_manifest_local.json --catalogs-config /tmp/catalogs_config_v2.json --locator-inventory sealed_2026-08-05/locator_inventory.jsonl --out-dir results_2026-08-07/phase3/recovery_benchmark | local-cpu | free (local) | <5min (given cached VizieR reference classes) | runnable now |
Ledger #6 first discriminator - local-PNG correction to the z=8-14 massive-galaxy abundance at f_NL = -35/16 vs -35/8 vs 0 (LoVerde+2008 Edgeworth mass function) anomaly-map-png-highz-abundance | f_NL = -35/16 matter-contraction squeezed value (ledger #1, closed in-lab) | python3 research/anomaly_map/ledger6_png_highz_abundance.py | local | free (local) | ~3 s | runnable now |
Papers in this program
| Work | Code | Role | Paper |
|---|---|---|---|
| Rebuilt DESI anomaly-science flagship (future primary paper) | anomaly-flagship | lead | — |
| DESI Public-ID Recovery Catalog (supporting release) | P3-support | support | view |
External data sources
| Source | Kind | License | Link |
|---|---|---|---|
| DESI DR1 iron zcatalog (zall-pix-iron.fits) | dataset | — | source ↗ |
| DESI DR1 (base release + coadd corpus) | dataset | — | source ↗ |
| SPARCL (spectra retrieval API) | api | — | source ↗ |
| NEOWISE IR variability catalog (IRSA) | dataset | — | source ↗ |
| NANOGrav 15-yr dataset | dataset | — | source ↗ |
| bamfai/bigbounce-anomaly-catalog (HF) | dataset | — | source ↗ |
| GitHub mirror (Hubify-Projects/bigbounce) | code-mirror | — | source ↗ |
| P3 Zenodo archive | zenodo-archive | — | source ↗ |
Research program · bounce-theory
Bounce theory
Question: Does matter-dominated contraction produce a distinctive, reproducible primordial non-Gaussian amplitude?
54 experiments · 52runnable now · full reproduction $10.30 · ~1-2 days sequential; dominated by the P1A delta-Neff MCMC (~12-18h to R-1<0.01) and the P1B SN-overlap control chains (several hours per chain pair). All 6 P2 derivation/analysis scripts are minutes-scale and independently parallelizable; the 500-MC NaMaster birefringence recovery (~1.5-2h) and NaMaster window regen (~5-15 min) can run alongside the MCMC legs.
Experiments — reproduction order
| Experiment | Inputs | Scripts | Venue | Est. cost | Wall-clock | State |
|---|---|---|---|---|---|---|
Four-vertex f_NL^local = -35/16 amplitude derivation (P2 headline) p2-vertex-check | internal derivation — no external input | python3 p2_vertex_check.py python3 fig_4vertex_sum.py python3 exact_shape_analysis.py | local | free (local) | minutes | runnable now |
G1 gradient-transmission scheme-dependence (Phase 1, T_c ~ 1/dcut) p2-g1-gradient-transmission · depends on p2-vertex-check | internal derivation — no external input | python3 g1_gradient_transmission_scheme.py | local | free (local) | minutes | runnable now |
G1 dressed-metric (Wilson-Ewing) transmission closure (T_c(k)=1, |delta f_NL| <= 6.8e-8) p2-g1-dressedmetric-transmission · depends on p2-g1-gradient-transmission | internal derivation — no external input | python3 g1_dressedmetric_transmission.py python3 g1_dressedmetric_ic_close.py | local | free (local) | minutes | runnable now |
Track A2: transmission coefficient of the matter-contraction f_NL through explicit nonsingular bounces (linear-transfer term, scheme-labeled) p2-a2-bounce-fnl-transmission · depends on p2-g1-dressedmetric-transmission | internal derivation — no external input | python3 a2_transmission_linear.py python3 a2_transmission_figures.py | local | free (local) | under 10 s total | runnable now |
Track A2 lane (a): cubic-vertex table for zeta through a nonsingular bounce — coefficient poles at H=0, scheme regularisation, super-Hubble in-in reduction and S1 bounce-window estimate p2-a2-lane-a-cubic-vertex-table · depends on p2-a2-bounce-fnl-transmission, p2-fnl-adjudication-inin-from-scratch | a2_transmission_linear backgrounds | python3 cubic_vertex_table.py | local | free (local) | under 10 s | runnable now |
Track A2 lane (b): numerical bounce-window in-in evaluation of Delta f_NL^bounce on the three A2 backgrounds — every vertex V1-V7 plus redefinition terms, scheme S1, with eta_*/window/step convergence tests and the S2 divergence diagnostic p2-a2-lane-b-numerical-inin · depends on p2-a2-lane-a-cubic-vertex-table, p2-a2-bounce-fnl-transmission | a2_transmission_linear backgrounds + adiabatic-vacuum mode evolution lane (a) vertex table + regularisation prescription (S1/S2, pole orders, closed form -(5/24) rho_B) | python3 bounce_cubic_inin.py | local | free (local) | under 15 s | runnable now |
Track A3 (ledger row 9 / A3-1e) lane 9b: is the effective-fluid (S2) divergence of Delta f_NL^bounce physical? Exact S2 modes at H=0, regularity of the comoving-gauge constraint solutions, total-derivative origin of the Maldacena-form poles p2-a3-lane-9b-s2-regularisation · depends on p2-a2-lane-a-cubic-vertex-table, p2-a2-lane-b-numerical-inin | lane (a) cubic-vertex table (coefficients, S2 pole counting) lane (b) S2 divergence report (d_cut slope -1.005..-1.007) | python3 lane9b_s2_regulation.py | local | free (local) | under 10 s | runnable now |
Track A3 (ledger row 9 / A3-1e) lane 9b-2: scheme-S2 (effective-fluid MS variable) bounce-window cubic contribution from the RAW ADM cubic Lagrangian on exact S2 modes, end-to-end in-in through the Quintin-type bounce, compared with scheme S1's -(5/24) rho_B p2-a3-lane-9b2-s2-rawadm · depends on p2-a3-lane-9b-s2-regularisation, p2-a2-lane-b-numerical-inin | lane 9b finding (S2 divergence = total-derivative pole; raw ADM form finite on exact S2 modes; Frobenius data) lane (b) S1 numbers, in-in conventions, window/eta* tests (Quintin rho_B = 0.6700, -(5/24) rho_B = -0.1396, f_NL^after = -0.5002) lane (b) in-in engine (conventions reproduced: -2 Im commutator, 3! attachments once, f_NL = (5/6) B / sum PP) adjudicated matter-contraction value f_NL = -35/16 (ledger #1; used as an independent gate, not an input) | python3 lane9b2_s2_rawadm.py | local | free (local) | under 1 min | runnable now |
Track A3 (ledger row 18a / A3-S2r): tensor transfer through the Quintin-type bounce and the like-for-like post-bounce tensor-to-scalar ratio r_after per scalar continuation scheme (S1 geometric z=a vs S2 effective-fluid MS variable), including the c_s dependence p2-a3-row18a-s2-tensor-transfer · depends on p2-a3-lane-9b2-s2-rawadm | lane 9b-2 exact S2 scalar modes, Quintin-type background, S1 reference rho_B = 0.6699892 (imported as a module; single source of truth for the background and both scalar schemes) row 10 tensor/scalar transfer on the poly and LQC backgrounds (T_h = T_zeta[S1] to 8.5e-9, r_before = r_after = 24) used as an independent cross-check, not as an input | python3 row18a_s2_tensor.py | local | free (local) | under 10 s | runnable now |
Ledger row 9 (A3-1e) lane (a): does the Quintin+2015 scalar-field-velocity-dip amplification of zeta (their Eq. 79) exist on the lab's three A2 backgrounds, and what is the curvature-spectrum transfer in the band k eta_B in [0.1, 10] that the S1 super-Hubble result does not cover p2-a3-lane-9a-velocity-dip · depends on p2-a2-bounce-fnl-transmission, p2-a2-lane-b-numerical-inin | a2_transmission_linear backgrounds (Quintin-type, LQC effective dust, poly non-LQC) + adiabatic-vacuum mode evolution + exact S/C matter-basis projection lane (b) numerical in-in Delta f_NL^bounce (the term this lane's transfer would modulate) | python3 lane9a_velocity_dip.py | local | free (local) | under 4 min | runnable now |
G3 torsion four-fermion bound (Einstein-Cartan estimate) p2-g3-torsion-fourfermion-bound · depends on p2-vertex-check | internal derivation — no external input | python3 g3_torsion_fourfermion_bound.py | local | free (local) | minutes | runnable now |
Honest-negative in-in bounce attempts (pathz / pathz2, superseded) p2-honest-negative-inin | internal derivation — no external input | python3 pathz_full_inin_bounce.py python3 pathz2_calibrated_inin.py | local | free (local) | minutes | superseded |
Channel-native Fisher surrogate (c15) + covariance chain (c8-c15) p2-channel-native-fisher · depends on p2-vertex-check | CAMB theory spectra Heinrich et al. 2023 SPHEREx covariance (Cov_B) | python3 c8_fnl_running_fisher.py python3 c15_channel_native_fisher.py python3 c10_joint_covariance_marginalization.py python3 p2_joint_cov.py | local | free (local) | 2-6 hours (CAMB-bound) | needs data restore |
MCMC full-tension / Planck+BAO+SN / third-combo chains (delta N_eff) p1a-mcmc-dneff | Cobaya full-tension config Cobaya Planck+BAO+SN config Cobaya Planck config Cobaya Planck+BAO config | cobaya-run cobaya_full_tension.yaml cobaya-run cobaya_planck_bao_sn.yaml cobaya-run cobaya_planck.yaml cobaya-run cobaya_planck_bao.yaml | runpod (32-vCPU CPU-optimized instance) | $8 | 12-18 hours to R-1<0.01 convergence | runnable now |
500-MC NaMaster EB birefringence recovery (beta = 0.238 deg, SNR 20.3) p1a-namaster-500mc-birefringence | internal derivation — no external input | python3 physical_spectra.py (checkpoint/resume harness) | runpod (CPU-bound instance, ~$0.17/hr class) | $0.30 | ~1.5-2h for 500 MC realizations at nside=512, lmax=1024 | runnable now |
100,000-sample N_tot sensitivity Monte Carlo (Spearman |rho_s|=0.996) p1a-ntot-sensitivity-mc | internal derivation — no external input | python3 vacuum_scale_sensitivity_scan.py | local | free (local) | minutes (100,000-sample Monte Carlo is CPU-light) | runnable now |
ALP prior-predictive / spectator-conditioned prior-predictive p1a-alp-prior-predictive | internal derivation — no external input | python3 alp_prior_predictive.py python3 alp_spectator_conditioned_prior_predictive.py | local | free (local) | minutes | runnable now |
SN-overlap control chains A (Pantheon+) / B (DES-SN5YR) p1b-sn-overlap-control-chains · depends on p1a-mcmc-dneff | Pantheon+ supernova compilation DES-SN5YR supernova compilation | cobaya-run cobaya_control_pantheonplus.yaml cobaya-run cobaya_control_desy5.yaml | runpod (A4000-class, $0.17/hr) | $2 | several hours per chain pair to R-1<0.01 | runnable now |
NaMaster window regenerability check (pymaster 3.0) p1b-namaster-window-regen · depends on p1a-namaster-500mc-birefringence | internal derivation — no external input | python3 rebuild_workspace_check.py | local or runpod (CPU-only, no GPU needed) | free (local) | ~5-15 minutes | runnable now |
Adjudication of the matter-contraction local f_NL: from-scratch in-in (validated on de Sitter and USR), delta-N on both slicings, long-mode shear — NEXT_SCIENCE_LEDGER #1 closure p2-fnl-adjudication-inin-from-scratch | lab second-method delta-N (uniform-density slices) — used for: separate-universe system and the -55/16 result, reproduced here for general eps and compared with the comoving-slice delta-N BigBounce Paper 2 Appendix A + scripts/p2_vertex_check.py — used for: per-vertex table tab:vertexwalk and Eq. vertexsum, reproduced exactly; 'spurious term' narrative corrected | python3 research/theory_audit/fnl_matter_contraction_adjudication_2026_09_02.py | local | free (local) | 2-5 minutes (exact sympy; eight parallel vertex integrations) | runnable now |
Independent second-method matter-contraction f_NL (separate-universe / nonlinear delta-N) — NEXT_SCIENCE_LEDGER #1 p2-fnl-second-method-deltan | BigBounce Paper 2 manuscript — Appendix A four-vertex re-summation (-35/16) for the reconciliation table | python3 research/theory_audit/fnl_matter_contraction_second_method_2026_09_02.py | local | free (local) | under 1 minute | runnable now |
A3-2: second-order Bianchi-I (anisotropic) separate-universe cross-check of the matter-contraction squeezed f_NL, including the long mode's shear a3-2-fnl-bianchi-separate-universe · depends on p2-fnl-adjudication-inin-from-scratch, p2-fnl-second-method-deltan | lab in-in adjudication (2026-09-02) lab second-method delta-N (2026-09-02) | python3 research/theory_audit/fnl_bianchi_separate_universe_2026_09_03.py | local | free (local) | under 1 minute | runnable now |
Adjudication of the squeezed-limit monopole of the matter-contraction f_NL: in-in (-15/8) vs comoving delta-N (-5) via the classical O(k^0) super-Hubble kernel with L/K/X tagging p2-fnl-monopole-adjudication · depends on p2-fnl-adjudication-inin-from-scratch, p2-fnl-second-method-deltan, a3-2-fnl-bianchi-separate-universe | lab in-in adjudication (2026-09-02) lab second-method delta-N (2026-09-02) lab Bianchi-I separate universe (2026-09-03) | python3 research/theory_audit/fnl_monopole_adjudication_2026_09_03_general_eps.py && python3 research/theory_audit/fnl_monopole_adjudication_2026_09_03.py python3 research/theory_audit/fnl_monopole_adjudication_2026_09_03_general_eps.py | local | free (local) | under 1 minute | runnable now |
Track A3 item A3-1 — compaction-function PBH abundance with local non-Gaussianity at f_NL = -35/16 vs -35/8 vs 0 (Choudhury et al. 2025 formalism) a3-pbh-compaction-fnl | internal derivation — no external input | python3 research/track_a3_multichannel/pbh_compaction_fnl.py | local | free (local) | ~4 min | runnable now |
Track A3 channel 2 — Press-Schechter PBH abundance with local quadratic non-Gaussianity at f_NL = -35/16 vs -35/8 vs 0 a3-pbh-abundance-fnl | internal derivation — no external input | python3 research/track_a3_multichannel/pbh_abundance_fnl.py | local | free (local) | 0.03 s (measured) | runnable now |
Track A3 channel 1 — reproduction of the NANOGrav 15-yr free-spectrum gamma posterior and Savage-Dickey Bayes factors from the committed chain a3-pta-gamma-reproduction | NANOGrav 15-yr HD-correlated free-spectrum emcee chain (320,000 samples) committed reference summaries for the diff (results.json + savage_dickey_2026-05-29.json in this dir) | python3 research/track_a3_multichannel/pta_gamma_reproduce.py | local | free (local) | 0.02 s (measured) | runnable now |
Track A3 §IV C closure (DA3M-R2-01) — injection-recovery test at gamma=13/3 and gamma=3 through the same 30-bin free-spectrum interpolated-density likelihood/priors as emcee_freespec.py a3-pta-injection-30bin-2026-09-02 | internal derivation — no external input | python3 research/track_a3_multichannel/pta_injection_30bin_2026_09_02.py | local | free (local) | 6 s (measured) | runnable now |
Track A3 channel 3 — survey reach and current-constraint tension table for f_NL^local = -35/16 a3-survey-reach-fnl | published sigma(f_NL) forecasts and DESI DR1 measurement (literature values, quoted verbatim from abstracts) | python3 research/track_a3_multichannel/survey_reach_fnl.py | local | free (local) | 0.01 s (measured) | runnable now |
A3-1b - the lab's own matter-bounce curvature power spectrum Delta^2_zeta(k) at PBH scales: delivered vs required amplitude in the compaction-function PBH criterion, and the COBE/FIRAS mu-distortion check on the early-SMBH seed amplitude a3-1b-inlab-delta2-zeta · depends on a3-pbh-compaction-fnl, p2-a2-bounce-fnl-transmission | A3-1 compaction-function PBH machinery (imported unmodified; only its module-level spectrum function is swapped and restored) A2 linear bounce transmission - eta_B per background, scale-independence of the transfer for k eta_B << 1 | cd research/track_a3_multichannel && python3 inlab_delta2_zeta_2026-09-03.py | local | free (local) | ~3 min | runnable now |
A3-3 - scalar-induced gravitational waves in the NANOGrav band from the lab's own curvature power spectrum: does the CMB-anchored matter-bounce spectrum give the paper's Channel-I gamma = 3? a3-3-sigw-nhz-from-lab-spectrum · depends on a3-1b-inlab-delta2-zeta, p2-a2-bounce-fnl-transmission, a3-pta-gamma-reproduction | A3-1b lab curvature spectrum Delta^2_zeta(k) - the exact spectrum evaluated here A2 linear bounce transmission - eta_B per background; scale-independence of the transfer for k eta_B << 1 (validity test in section 3 of the note) The paper's own gamma convention Omega_GW ~ f^{5-gamma} | cd research/track_a3_multichannel && python3 sigw_nhz_from_lab_spectrum_2026_09_04.py | local | free (local) | ~5 s | runnable now |
A3M R5-15 - the model's own FIRST-ORDER (primordial) tensor Omega_GW h^2 in the NANOGrav band, compared with the scalar-induced background and with NANOGrav: which dominates, and does the PTA null change? a3-r5-15-tensor-omega-nhz · depends on a3-3-sigw-nhz-from-lab-spectrum, a3-pta-gamma-reproduction | A3-3 induced (second-order) background: Omega_GW h^2(f_yr) = 1.4545e-23 (anchored) / 5.8764e-23 (dust) - the comparator this item is measured against, and the source of the identical f<->k map and transfer prefactor The paper's gamma convention Omega_GW ~ f^{5-gamma} and its 'prim. tensors n_T = 0' Table II row | cd research/track_a3_multichannel && python3 r5_15_tensor_omega_nhz.py | local | free (local) | ~1 s | runnable now |
Ledger row 10 (A3-4 + A3-ns) - the matter bounce's OWN tensor-to-scalar ratio r and scalar tilt n_s, derived for the dust contraction and propagated through all three A2 bounce backgrounds a3-4-row10-r-ns · depends on p2-a2-bounce-fnl-transmission, a3-r5-15-tensor-omega-nhz, a3-1b-inlab-delta2-zeta | A2 linear-transmission module - the three bounce backgrounds (LQC-effective dust, poly-analytic non-LQC, Quintin2015-type), the adiabatic-vacuum scalar evolution and the exact matter-basis projection, imported directly First-order tensor Omega_GW at nHz - re-run with the model's own r as CASE C | cd research/track_a3_multichannel/row10_r_ns && python3 row10_r_ns.py | local | free (local) | ~3 s | runnable now |
A3M R5-18 - gamma_cr coverage of the 27-point PBH (Delta, r_p k_p, C_th) grid: how many points sit on the enhancement branch (gamma_cr <= 0.85), and is the quoted 1.7-1.9 required-amplitude ratio inside the scanned coverage? a3-r5-18-gammacr-coverage · depends on a3-pbh-compaction-fnl, a3-1b-inlab-delta2-zeta | A3-1 compaction-function PBH grid - the 27 (Delta, r_p k_p, C_th) points with gamma_cr and ratio_-35/16_over_-35/8 (robust_amplitude_requirement_grid) A3-1b in-lab curvature spectrum - gamma_cr of the lab's own near-scale-invariant shape over the IR-cutoff scan (ir_cutoff_sensitivity) A3-1b note - the 1.85-1.89 ratio at the in-lab shape and the observation that it lies outside the grid's [1.610, 1.809] A3-1 note - the sign-flip statement (enhancement at gamma_cr <~ 0.85, suppression above), step (4) of the generating script | cd research/track_a3_multichannel && python3 r5_18_gammacr_coverage.py | local | free (local) | ~1 s | runnable now |
Ledger row 9 (A3-1e) lane (c) - literature-bound map of the Agullo-Bolliet-Sreenath 2017 LQC third-order Hamiltonian onto the lab's classical scheme-S1 cubic vertex table, plus the k_LQC -> k*eta_B scale-window conversion and PTA/PBH overlap test ledger9-c-abs-operator-map · depends on p2-a2-lane-a-cubic-vertex-table, a3-3-sigw-nhz-from-lab-spectrum, a3-1b-inlab-delta2-zeta | Lane (a) cubic vertex table V1-V7 + R1-R4 and the LQC-dust background Lane (c) 2026-09-03 literature comparison (the statement corrected by this lane) A3-3 SIGW note - comoving bounce scale k_B = 1.71e15 Mpc^-1 at T_B = 1e8 GeV, linear in T_B A3-1b in-lab Delta^2_zeta - PBH-band amplitude 5e-10 to 1.3e-9 and the 7.0 dex deficit | python3 research/cubic_bounce_transmission/lane9c_abs_operator/lane9c_scale_window.py | local | free (local) | ~2 s (script); ~2 min to re-fetch and re-read the arXiv source | runnable now |
Ledger #4 step 1 - provenance-bound download and lab-native sanity check of the DESI DR1 public LSS clustering catalogues (QSO, LSScats v1.5), the input products for an independent reproduction of the DR1 local-PNG scale-dependent-bias constraint ledger4-desi-dr1-lss-sanity · depends on a3-survey-reach-fnl | Ledger #3 survey-reach study (DESI DR1 = 0.16 sigma on f_NL = -35/16; sets this item's honest scope) | python3 research/desi_png_reproduction/dr1_lss_sanity.py | local | free (local) | ~2 min including download | runnable now |
Ledger #7 gate - net helicity asymmetry Delta_h of the SGWB from the minimal Einstein-Cartan-Holst torsion bounce (symbolic parity-operator check + super-Hubble k-odd estimate) ledger7-chiral-gw-delta-h | On-shell ECH torsion irreps and O4 coefficient (in-lab adjudication, 2026-08-08) ECH Note four-fermion contact term L_4psi = -(3 kappa/16) gamma^2/(1+gamma^2) (J5.J5) Branch M PGT bounce GW spectrum (amplitude/frequency gap) Branch Q sourced-parity screening (BRANCH_Q_WEAK) | python3 research/chiral_gw_gate/chiral_gw_delta_h.py | local | free (local) | ~3 s | runnable now |
Ledger row 9 (A3-1e) lane 9c-2 - scheme-S1 bounce-window in-in integral (V2-V7 + R1-R4) evaluated with EXACT dressed-metric mode functions on the LQC-dust background over k*eta_B in [0.1, 10], with the initial state varied (lab adiabatic contraction vacuum, ABS adiabatic-order-zero vacuum, 4th-order adiabatic vacuum), testing whether the Agullo-Bolliet-Sreenath 2017 f_NL enhancement near k*eta_B ~ 1 appears in the lab's model ledger9-c2-lqc-exact-modes-inin · depends on ledger9-c-abs-operator-map, p2-a2-lane-a-cubic-vertex-table, a3-1b-inlab-delta2-zeta | Lane (a) cubic vertex table V1-V7 + R1-R4 (scheme S1 coefficients and kernels) Lane (b) numerical in-in machinery and its LQC result at k*eta_B = 1e-3 (-0.1043113297), the gate target of this lane Lane (b) results.json (backgrounds.lqc.k_scan, read by the gate at run time) A2 linear-transmission module - LQC-dust background, dressed geometric potential a''/a = x^(1/3)(1/6 + x/3), matter-basis projection, exact dust adiabatic vacuum Lane 9c - ABS operator map, k_LQC*eta_B = 1.060146, decay 1.830229 per k*eta_B, 408 sigma PBH tail A3-1b in-lab Delta^2_zeta ~ 1e-9 and the 7.0 dex PBH-band deficit | python3 research/cubic_bounce_transmission/lane9c2_lqc_modes/lane9c2_lqc_modes.py | local | free (local) | ~3 min | runnable now |
A3M ledger row 11(a) - operator-by-operator location of the Choudhury et al. 2025 compaction-function f_NL sign disagreement: does negative local f_NL suppress or enhance the PBH abundance at fixed Gaussian amplitude? a3-row11a-choudhury-sign · depends on a3-pbh-compaction-fnl, a3-r5-18-gammacr-coverage, a3-1b-inlab-delta2-zeta | A3-1 compaction-function integrator (Eqs. 52-54 covariances, Eq. 60 abundance) - imported unmodified as the numerical engine A3-1 note - the enhancement-at-gamma_cr<~0.85 statement (sec 4.3) this item adjudicates R5-18 gamma_cr coverage note - the in-lab shape's gamma_cr in [0.267, 0.630] and the standing of the quoted 1.7-1.9 ratio | cd research/track_a3_multichannel/row11_pbh_residuals && python3 row11_choudhury_sign.py | local | free (local) | ~10 s | runnable now |
Row 11(c): second-order threading map from Maldacena's comoving zeta to the zero-shift (fluid-congruence) delta N_c in a non-attractor contraction; mechanism behind the delta N = -5 vs in-in -15/8 monopole gap row11c-threading-map-second-order · depends on p2-fnl-monopole-adjudication | lab monopole adjudication (2026-09-03) lab monopole adjudication script (conventions) | python3 research/theory_audit/threading_map_second_order_2026_09_04.py | local | free (local) | under 5 minutes | runnable now |
A3M ledger row 11(b) - compaction-function scan extended to gamma_cr in [0.2, 1.0]: the required-amplitude ratio A(-35/16)/A(-35/8) INSIDE the in-lab spectrum shape's own coverage [0.267, 0.630], and whether the quoted '1.7-1.9' survives a3-row11b-gammacr-extension · depends on a3-pbh-compaction-fnl, a3-r5-18-gammacr-coverage, a3-row11a-choudhury-sign | A3-1 compaction-function integrator (Eqs. 52-54 covariances, Eq. 60 abundance) - imported unmodified as the numerical engine A3-1 note - the enhancement-at-gamma_cr<~0.85 statement (sec 4.3) this item adjudicates R5-18 gamma_cr coverage note - the in-lab shape's gamma_cr in [0.267, 0.630] and the standing of the quoted 1.7-1.9 ratio Row 11(a) sign adjudication - establishes that gamma_cr controls the sign of the f_NL response, hence that the ratio must be quoted inside the in-lab gamma_cr coverage | cd research/track_a3_multichannel/row11_pbh_residuals && python3 row11_gammacr_extension.py | local | free (local) | ~30 min | runnable now |
Ledger row 14 - joint dependence of the tensor-to-scalar ratio r and the squeezed local f_NL on the contraction's scalar sound speed c_s, and the (empty) window in which r < 0.036 with an acceptable f_NL a3-row14-cs-window · depends on a3-4-row10-r-ns, p2-fnl-adjudication-inin-from-scratch | A2 linear-transmission module - the three bounce backgrounds and the adiabatic-vacuum scalar evolution with exact matter-basis projection, imported directly Lab in-in adjudication of the matter-contraction f_NL at c_s = 1 (comoving-gauge isoceles squeezed limit -35/16), the c_s -> 1 cross-check target Ledger row 10 - r = 16 epsilon = 24 at c_s = 1, bounce-invariant; the starting point of this row | cd research/track_a3_multichannel/row14_cs_window && python3 row14_cs_window.py | local | free (local) | ~6 s | runnable now |
Ledger row 18(b) - c_s-dependence of the bounce's own cubic contribution Delta f_NL^bounce(c_s) in scheme S1, and the resulting shift of the joint (r, f_NL) window boundary of row 14 a3-row18b-cs-bounce-cubic · depends on p2-a2-lane-b-numerical-inin, p2-a2-lane-a-cubic-vertex-table, a3-row14-cs-window | Lane (a) cubic-vertex table - the c_s-dependent coefficients c_V(a,H,eps,eta_sr,c_s) of the P(X,phi) cubic action (Chen, Huang, Kachru & Shiu 2007; Seery & Lidsey 2005; Maldacena 2003) transcribed and checked Lane (b) S1 in-in integrator at c_s = 1 - imported directly for its vertex slot/kernel definitions, mode class, dot-product table and quadrature; supplies the c_s = 1 regression gate (-0.139818 / -0.104311 / -0.127111) A2 linear-transmission module - the three bounce backgrounds (Quintin-type, LQC-effective dust, poly) and the adiabatic-vacuum scalar evolution, imported directly Ledger row 14 - f_NL^pre(c_s) = -165/16 + 65/(8 c_s^2), r = 24 c_s, and the c_s-independence of the bounce transfer T (verified to 4e-11); the window this row re-evaluates | cd research/cubic_bounce_transmission/row18b_cs_bounce_cubic && python3 row18b_cs_bounce_cubic.py | local | free (local) | ~10 s | runnable now |
Ledger row 19 - the joint (r, f_NL) no-go for general P(X) k-essence with the cubic-action coefficient lambda free: lambda-general f_NL^pre(c_s, lambda/Sigma), the exact lambda-independence of r and of the bounce's own cubic term, and the window scan a3-row19-lambda · depends on a3-row14-cs-window, a3-row18b-cs-bounce-cubic, p2-a2-lane-a-cubic-vertex-table | Ledger row 14 - f_NL^pre(c_s) = -165/16 + 65/(8c_s^2) at lambda = 0, r = 24 c_s, and the transfer T = 0.16500538 (Quintin background); the lambda = 0 result this row generalises Ledger row 18(b) - Delta f_NL^bounce(c_s) = -(5/24) rho_B (6c_s^2-5)/c_s^4 in scheme S1; its integrator is imported and its V1 coefficient patched to carry lambda Lane (a) cubic-vertex table - c_V1 = -a^3[Sigma(1-1/c_s^2) + 2 lambda]/H, the lambda-carrying vertex Lab in-in adjudication - the canonical-field squeezed value -35/16 that the lambda-general formula must reproduce at c_s = 1, lambda = 0 | cd research/track_a3_multichannel/row19_lambda && python3 row19_lambda.py | local | free (local) | ~5 s | runnable now |
Ledger row 15 - the curvaton-type matter bounce: whether a light spectator in the dust contraction gives r < 0.036 with n_s ~ 0.965 and an O(1) local f_NL, and what happens to the intrinsic -35/16 at that point a3-row15-curvaton · depends on a3-row14-cs-window, a3-4-row10-r-ns | A2 linear-transmission module - the three bounce backgrounds and the u'' + (k^2 - a''/a)u = 0 evolution, imported directly for the spectator's frozen-branch transfer Ledger row 10 - r = 16 epsilon = 24 and n_s - 1 = 12w/(1+3w) on the w = -0.0029 anchor; the tensor amplitude and tilt this row reuses unchanged Ledger row 14 - the c_s no-go and the A2 transfer values T = 0.165-0.250 quoted here Lab in-in adjudication of the matter-contraction f_NL (-35/16), the amplitude diluted by (r/24)^2 in this row Branch-W ALP-curvaton tilt program, phase 1 - the lab's existing curvaton work; its n_sigma - 1 sign is corrected here | cd research/track_a3_multichannel/row15_curvaton && python3 row15_curvaton.py | local | free (local) | ~6 s | runnable now |
Row 15 adjudication: independent sympy re-derivation of the curvaton-type matter bounce claims (CXB11 Eq. 18 coefficient and Case-1 f_NL, branch-W tilt sign, spectator MS operator and tilt, two-channel r, delta-N curvaton f_NL, adiabatic-bispectrum dilution) a3-row15-curvaton-adjudication · depends on a3-row15-curvaton | row-15 lane under adjudication branch-W ALP-curvaton tilt note | python3 research/theory_audit/curvaton_matter_bounce_adjudication_2026_09_04.py | local | free (local) | under 1 minute | runnable now |
Ledger row 15b - the entropy (spectator) sector in the three A2 matter-bounce backgrounds: lambda_sigma vs lambda_T vs lambda_zeta per scheme, and the pre-bounce condition on r_dec Mpl/sigma_* for r < 0.036 a3-row15b-entropy-sector · depends on a3-row15-curvaton, p2-a3-row18a-s2-tensor-transfer | A2 linear-transmission module - the three bounce backgrounds (Quintin-type, LQC-effective-dust, poly-analytic) and the a''/a arrays used for both the spectator and the tensor evolution Ledger row 18a - lambda_T, lambda_zeta^S1 and lambda_zeta^S2 on the Quintin background; the S2 scalar transfer imported here Ledger row 15 - the curvaton window r = 24/[1 + (4/3) r_dec^2 (Mpl/sigma_*)^2], the F >= 25.82 requirement and the (r/24)^2 dilution of the intrinsic -35/16 Curvaton matter-bounce adjudication - confirms the spectator obeys the same MS operator for constant epsilon, so n_s is inherited | cd research/track_a3_multichannel/row15b_entropy_sector && python3 row15b_entropy_sector.py | local | free (local) | ~10 s | runnable now |
Novelty lift #2: criterion for the O(1) failure of the isotropic separate universe (delta N) in non-attractor phases — <eps/c_s^2>_zeta, general-w lambda and f_map, validations on dust / USR / attractor / ekpyrotic lift2-separate-universe-failure-criterion | lab threading map (2026-09-04), frozen constant-eps kernels lab monopole adjudication (2026-09-03), in-in general-eps input novelty audit item C2 | python3 research/theory_audit/separate_universe_failure_criterion_2026_09_04.py | local | free (local) | under 10 seconds | runnable now |
PSU R1 science gates S1/S2 (+S3 math): label-resolved second-order composition (initial vs final worldline label), well-defined separate-universe criterion with the gradient term restored, change-of-variable statement psu-gates-s1-s2-label-composition-criterion · depends on lift2-separate-universe-failure-criterion | PSU v1S.0.1 R1 truth audit (items S1, S2, S3) lab threading map (2026-09-04), frozen map pieces incl. lab_init / wl_initextra separate-universe failure criterion note (2026-09-04) lab monopole adjudication (2026-09-03), in-in general-eps kernel | python3 research/theory_audit/psu_gates_S1_S2_2026_09_04.py | local | free (local) | under 10 seconds | runnable now |
PSU R2 science gates S6-S11: map-independence sentinel + Eq.(4) assertion, Cai factor-2 scope, exact numerical USR delta N(phi,pi) at finite eps_s, final-slice dependence, constant-piece long mode, Zenodo status psu-gates-s6-s11-science-gates · depends on psu-gates-s1-s2-label-composition-criterion | PSU v1S.0.2 R2 truth audit, section 5(ii) science items S6-S11 lab threading map (2026-09-04), frozen map pieces PSU gates S1/S2 note (closed forms re-asserted here) lab monopole adjudication (2026-09-03), in-in kernel and Cai/Li discussion | S6_SENTINEL_JSON=research/theory_audit/psu_gates_S6_sentinel_threading_2026_09_04.json python3 research/theory_audit/psu_gates_S6_S11_2026_09_04.py | local | free (local) | under 3 minutes | runnable now |
PSU gate S7: equation-level location of the factor 2 between Cai et al. 2009 (0903.0631) Eq. (37) and Eqs. (38)-(41)/Fig. 5, and the Li et al. 2017 (1612.02036) c_s=1 correspondence psu-gate-s7-cai-factor-2 · depends on psu-gates-s6-s11-science-gates | PSU gates S6-S11 note, item S7 lab f_NL matter-contraction adjudication (claims under test) | python3 research/theory_audit/psu_gate_S7_cai_factor_2026_09_05.py | local | free (local) | under 1 minute | runnable now |
PSU gates S9 + S10 at second order: uniform-density-slice threading map (delta N_c,rho from the in-in bispectrum) and the constant-long-mode kernel K_c of the general-label map psu-gates-s9-s10-rho-slice-kc · depends on psu-gates-s6-s11-science-gates, row11c-threading-map-second-order | PSU gates S6-S11 (S9/S10 sections: linear rho-slice map, f_map(g) with K_c open) threading map at second order (machinery + frozen phi-slice kernels) lab separate-universe delta N on uniform density (comparison value 5(eps-7)/8 = -55/16) | python3 research/theory_audit/psu_gates_S9_S10_2026_09_05.py | local | free (local) | under 3 minutes | runnable now |
PSU gate S9b: intrinsic flat-slice initial-data non-Gaussianity through the delta-N map (does it close the 5(6-eps)/24 gap between -55/16 and -5/2?) psu-gate-s9b-intrinsic-initial-data-term · depends on p2-fnl-second-method-deltan, psu-gates-s9-s10-rho-slice-kc | PSU gates S9+S10 (S9.4 names the missing step) lab separate-universe delta N on uniform density (ODE, closed forms, -55/16) | python3 research/theory_audit/psu_gate_S9b_intrinsic_term_2026_09_05.py | local | free (local) | under 1 minute | runnable now |
PSU gate S9c: is the 5(6-eps)/24 gap between the delta-N uniform-density value -55/16 and the threading rho-slice value -5/2 the shift-divergence term dropped in the separate-universe evolution? psu-gate-s9c-evolution-residual-shift-term · depends on psu-gate-s9b-intrinsic-initial-data-term, psu-gates-s9-s10-rho-slice-kc, p2-fnl-second-method-deltan | PSU gate S9b (residual assigned to the evolution step) PSU gates S9+S10 json (S9.A2_superhubble_growing, f_map_rho tables) threading map at second order (worldline identity, eq. 2) lab separate-universe delta N on uniform density (-55/16) | python3 research/theory_audit/psu_gate_S9c_evolution_residual_2026_09_05.py | local | free (local) | under 1 minute | runnable now |
Adjudication of the L x S second-order lapse monopole A2(eps) on the uniform-density slice: S9 (eps(3-eps)^2/3, f^rho=-5/2) vs S9c (2(3-eps)^2 required, f^rho=-55/16) a2-lapse-monopole-adjudication · depends on psu-gate-s9c-evolution-residual-shift-term, psu-gates-s9-s10-rho-slice-kc | PSU gates S9+S10 (rho-slice map, A2 = eps(3-eps)^2/3, f^rho = 5(2eps-15)/24) PSU gates S9+S10 json (S9.A2_superhubble_growing) PSU gate S9c (exact separate universe: -5 comoving, 5(eps-7)/8 uniform density; A2 = 2(3-eps)^2 required) | python3 research/theory_audit/a2_lapse_monopole_adjudication_2026_09_07.py | local | free (local) | under 1 minute | runnable now |
Papers in this program
| Work | Code | Role | Paper |
|---|---|---|---|
| f_NL forecast / exact matter-contraction non-Gaussianity (PRD) | P2 | lead | view |
| Algebraic Cartan elimination (CQG Note) | P1A | support | view |
| namaster-proof: exact pseudo-Cl window inference and tamper-evident provenance (JORS) | P1B | support | view |
External data sources
| Source | Kind | License | Link |
|---|---|---|---|
| Planck PR4/PR3 CMB likelihoods (via Cobaya) | likelihood | — | source ↗ |
| BAO compilations (via Cobaya) | likelihood | — | source ↗ |
| Pantheon+ supernova compilation | dataset | — | source ↗ |
| DES-SN5YR supernova compilation | dataset | — | source ↗ |
| Heinrich et al. 2023 SPHEREx forecast covariance (Cov_B) — NOT publicly released by original authors; DP2-26/DP2-29 gap | covariance-matrix | — | not publicly released |
| P1A archive (algebraic Cartan elimination manuscript + data) | zenodo-archive | CC-BY-4.0 | source ↗ |
| P1B namaster-proof software archive | zenodo-archive | — | source ↗ |
| P1B namaster-proof paper archive | zenodo-archive | — | source ↗ |
Research program · galaxy-chirality
Galaxy chirality
Question: Is there a large-scale observed-label chirality dipole in the released DESI imaging catalog?
29 experiments · 25runnable now · full reproduction $18 · ~2-3 days sequential; dominated by the G1 ViT-Small retrain (~4h on RunPod A4000) and the e2e mirror-flip full-catalog inference (10.45h wall on RunPod A100, 16.9M inferences). The C1-C3 NaMaster null-test batch, G2/G3/G4 validations, and the A_95 dipole limit are each minutes-scale once the trained model and full-catalog inference exist and can run in parallel. P5's DESI crossmatch build (~hours, local CPU) gates its four downstream environment analyses plus the cosmic-web/DESIVAST and r-conf legs, which are independently parallelizable after that.
Experiments — reproduction order
| Experiment | Inputs | Scripts | Venue | Est. cost | Wall-clock | State |
|---|---|---|---|---|---|---|
v2 ViT-Small production training (26,616-object historical realization) p4-v2-vit-production-training | GZ1 CW/CCW human-vote labels (S3-hosted) Galaxy Zoo DESI morphology predictions (Walmsley 2023) | python3 train_chirality_v2.py | runpod (A100/H200-class GPU, ViT-Small image classifier training) | $5 | 3-6 hours for a 26,616-object realization at typical ViT-Small batch throughput | superseded |
G1 — regenerable ViT-Small retrain with manifest (supersedes historical v2 training) p4-g1-vit-retrain-manifest | GZ1 CW/CCW human-vote labels (S3-hosted) GZ×DESI crossmatch | python3 train_g1_manifest.py | runpod A4000 16GB on-demand | $1 | ~1.5h smoke test + ~4-5h full retrain | runnable now |
G1 CE-included full composition (826-vs-846 adjudication) p4-g1-ce-composition-assembly · depends on p4-g1-vit-retrain-manifest | G1 retrained checkpoint | python3 g1_ce_composition_assembly.py | local CPU | free (local) | minutes to ~1 hour for the 26,609-object composition/adjudication pass | runnable now |
G2 — training-disjoint held-out GZ1 validation p4-g2-disjoint-validation · depends on p4-g1-vit-retrain-manifest | G1 retrained checkpoint GZ1 CW/CCW human-vote labels (S3-hosted) | python3 g2_disjoint_validation_v1_0_266.py | runpod A4000 or equivalent 16GB GPU (inference-only) | $0.05 | ~6-10 minutes | runnable now |
G3 — joint estimator covariance (local bootstrap leg + RunPod MASTER-leg refinement) p4-g3-joint-estimator-covariance · depends on p4-g1-vit-retrain-manifest | internal derivation — no external input | python3 g3_joint_estimator_covariance.py python3 g3_joint_estimator_covariance_master_v2.py | local CPU for the base bootstrap leg; RunPod A4000 (or a local pymaster install) for the MASTER-leg refinement | $0.36 | ~10-15 min local-bootstrap leg + ~1h MASTER-leg | runnable now |
e2e mirror-flip full-catalog inference (8.47M galaxies x 2 passes) p4-e2e-mirror-flip · depends on p4-g1-vit-retrain-manifest | G1 retrained checkpoint | python3 e2e_mirror_flip_fullrun.py python3 e2e_mirror_flip_transfer_function.py | runpod A100 (or equivalent 40GB-class GPU) | $12.44 | ~10-11 hours for the full 192-shard, 16,949,062-inference run | runnable now |
G4 — per-pixel confusion + generative parity-null (monopole mechanism) p4-g4-monopole-mechanism-injection · depends on p4-e2e-mirror-flip | Banked e2e mirror-flip per-galaxy pair record | python3 g4_monopole_mechanism_injection.py | local CPU or RunPod aggregation-only instance (no GPU inference required, reuses banked e2e mirror-pair record) | free (local) | tens of minutes for per-pixel confusion + generative parity-null aggregation over the banked record | runnable now |
A_95^obs coverage-calibrated dipole injection upper limit p4-a95-dipole-injection-limit · depends on p4-e2e-mirror-flip | internal derivation — no external input | python3 a95_observed_label_upper_limit_v1_0_265.py | local CPU | free (local) | ~10 minutes | runnable now |
C1 — NaMaster fsky sweep (part of the C1/C2/C3 monopole/dipole-null batch) p4-c1-namaster-fsky-sweep | internal derivation — no external input | python3 c1_p1b_namaster_fsky_sweep.py bash launch_c123_pod.sh | runpod CPU-strong instance (12+ vCPU, no GPU required) | $0.20 | ~1-1.5 hours for the 2x500 MC fsky sweep (fsky~0.85 and fsky~0.65) | runnable now |
C2 — monopole/dipole N_all binomial null p4-c2-nall-binomial-null · depends on p4-e2e-mirror-flip | internal derivation — no external input | python3 c2_p4_nall_binomial_null.py bash launch_c123_pod.sh | runpod CPU-strong instance or local CPU (job is CPU-bound) | $0.02 | ~6-10 minutes | runnable now |
C3 — Wp (N_all vs N_spiral) fsky invariance null p4-c3-wp-invariance-fsky · depends on p4-e2e-mirror-flip | internal derivation — no external input | python3 c3_p4_wp_invariance_fsky.py bash launch_c123_pod.sh | runpod CPU-strong instance or local CPU (job is CPU-bound) | $0.02 | ~6-10 minutes | runnable now |
GZ1-only classifier retrain + dipole null (pseudo-label independence check) p4-gz1only-retrain-dipole-null | GZ1 CW/CCW human-vote labels (S3-hosted) | python3 run_dipole_gz1only_fullN.py python3 gz1_stratified_confusion.py python3 validate_p4_v1_0_244_claims.py | local CPU for the surviving dipole-null recomputation; RunPod A4000 16GB would be needed to redo the full GZ1-only classifier retrain from scratch | free (local) | ~10 min local for the dipole-null recomputation on the surviving JSON (per COMPUTE_CAMPAIGN's own note); full end-to-end retrain time not estimable without the lost training script | needs data restore |
Empirical b/a (axis-ratio) DR8 morphology cross-match p4-dr8-axis-ratio-crossmatch | ls_dr8.tractor (Legacy Survey DR8 Tractor catalog, via NOIRLab Astro Data Lab TAP) | python3 pull_dr8_final.py python3 pull_dr8_datalab.py python3 edge_on_contamination_metric.py | local CPU or any RunPod instance (TAP pull + metric computation are CPU-bound, network-bound on the TAP query) | free (local) | ~30-60 minutes depending on TAP query throughput for the full catalog cross-match | needs data restore |
Dipole analysis (8.47M full-catalog) p4-dipole-8m-fullcatalog | internal derivation — no external input | python3 run_dipole_8M.py | runpod A100/H200-class GPU for the full 8.47M-galaxy dipole computation | $10 | several hours, scaling with catalog size and TTA passes | needs data restore |
Row 16(i) — full-parent (8,474,531-galaxy) real-space chirality dipole using the exact P4' primary estimator p4p-row16i-full-parent-dipole · depends on p4-v2-vit-production-training, p4-a95-dipole-injection-limit | Full DESI Legacy DR8 chirality catalog (class_eq, production equivariant Z2-TTA classifier; already-committed, not re-run) row16(ii) N=20,000 injection-calibrated postprocess residual bias | imports build_projector() verbatim from pipelines/p2_chirality/generate_p4_primary_label_shuffle_strict_v1_0_257.py python3 pipelines/p4prime_chirality_test/full_parent/run_full_parent_dipole.py | local | free (local) | ~1 min | runnable now |
P4xDESI DR1 crossmatch + matched catalog build p5-desi-dr1-crossmatch-build · depends on p4-e2e-mirror-flip | P4 galaxy-chirality catalog | python3 01_fetch_p4_catalog.py python3 02_fetch_desi_dr1.py python3 03_crossmatch.py | local CPU (implied local/CPU-bound crossmatch — no venue was explicitly logged for the original run) | free (local) | ~1-3 hours for the DESI DR1 fetch + 2,232,212-row crossmatch, dominated by DESI DR1 download bandwidth | runnable now |
P5 redshift-dependence analysis of spiral chirality p5-redshift-analysis · depends on p5-desi-dr1-crossmatch-build | P5 matched chirality x DESI DR1 catalog | python3 05_analysis_redshift.py | local CPU | free (local) | minutes to tens of minutes over the ~2.23M-row matched catalog | runnable now |
P5 local-density-dependence analysis of spiral chirality p5-density-analysis · depends on p5-desi-dr1-crossmatch-build | P5 matched chirality x DESI DR1 catalog | python3 06_analysis_density.py | local CPU | free (local) | minutes to tens of minutes over the ~2.23M-row matched catalog | runnable now |
P5 HEALPix sky-map analysis of spiral chirality p5-healpix-analysis · depends on p5-desi-dr1-crossmatch-build | P5 matched chirality x DESI DR1 catalog | python3 07_analysis_healpix.py | local CPU | free (local) | minutes to tens of minutes over the ~2.23M-row matched catalog | runnable now |
P5 systematics analysis of spiral chirality measurement p5-systematics-analysis · depends on p5-desi-dr1-crossmatch-build | P5 matched chirality x DESI DR1 catalog | python3 09_systematics.py | local CPU | free (local) | minutes to tens of minutes over the ~2.23M-row matched catalog | runnable now |
Cosmic-web / DESIVAST void analysis (16, 27, 35-39 series) p5-cosmic-web-desivast-void · depends on p5-desi-dr1-crossmatch-build | P5 matched chirality x DESI DR1 catalog | python3 16_cosmic_web_zshell_corrected.py python3 08_analysis_cosmic_web.py python3 27_rsd_void_recon_bound.py python3 35_desivast_cluster_bootstrap.py python3 36_desivast_native_selection_control.py | local CPU (no RunPod hit found for the p5 script family per the inventory's own grep) | free (local) | ~1-2 hours across the cluster-bootstrap and RSD/void-reconstruction steps over DESIVAST VAC + the ~2.23M-row matched catalog | runnable now |
r23conf/r24conf/r27conf closure recomputes p5-rconf-closures · depends on p5-desi-dr1-crossmatch-build | P5 matched chirality x DESI DR1 catalog | python3 21_r23conf_meta_closures.py python3 22_r24conf_local_batch.py python3 24_r24conf_pod_session.py python3 26_r27conf_ess_recomputes.py | local CPU (the pod-implying filename is unconfirmed; a local or RunPod CPU-strong instance both work) | free (local) | ~30 min - 2 hours across the four closure/batch scripts | runnable now |
Focal cluster inference sensitivity + interaction clustering robustness p5-focal-cluster-robustness · depends on p5-rconf-closures | P5 matched chirality x DESI DR1 catalog | python3 38_focal_cluster_inference_sensitivity.py python3 39_focal_interaction_clustering_robustness.py | local CPU | free (local) | ~30 min - 1 hour across both robustness scripts | runnable now |
Astra per-object crossmatch + HuggingFace mirror p5-astra-crossmatch-hf-mirror · depends on p5-desi-dr1-crossmatch-build | P5 matched chirality x DESI DR1 catalog | python3 15_astra_per_object_crossmatch.py python3 mirror_astra_to_hf.py | local CPU | free (local) | ~30 min - 1 hour for the per-object crossmatch + HF mirror push | runnable now |
P4' R2 closure (DP4P-22) — genuine 95% CL upper limit on the primary real-space chirality dipole amplitude via Neyman inversion of the injection-recovery null p4prime-a95-neyman-cl-2026-09-02 | Strict-primary catalog (887,472-galaxy / 23,633-pixel support) Committed fixed-occupancy 10^4-draw detection null Committed strict-primary generator (estimator/selection, imported verbatim) Committed detection-power A_95^obs script (injection model + estimator, imported verbatim) | python3 research/bh_universe_dipole/a95_upper_limit_2026_09_02.py | local | free (local) | ~80 s | runnable now |
P4' (Track C1) — confront the DESI chirality catalog's coverage-calibrated observed-label 95% sensitivity floor with Poplawski's rotating-black-hole-universe spin-axis claim p4prime-bh-universe-dipole-exclusion · depends on p4prime-a95-neyman-cl-2026-09-02 | A_95^obs and N_support (verbatim, not re-derived) Literature spin-axis amplitude claims (as published, cited by arXiv id, no re-analysis) Poplawski black-hole-universe mechanism and preferred-axis papers | python3 research/bh_universe_dipole/poplawski_dipole_exclusion_2026_09_02.py | local | free (local) | < 0.1 s | runnable now |
Row 16 (iv) — chirality x structure: parity vs environment, anomaly positions, redshift, and preferred axes row16iv-chirality-structure | P4 chirality catalog (primary-safe, DESI Legacy DR8 ViT+TTA, 8,474,531 rows) Anomaly catalog v2 science targets (1,244 rows) DESI spec-z x chirality crossmatch (P5) | imported python3 chirality_structure_env_z.py python3 chirality_structure_anomaly.py python3 chirality_structure_axes.py python3 chirality_structure_figures.py | local CPU | free (local) | ~1 hour | runnable now |
Row 16(i-b) — is the full-parent chirality dipole a QC/footprint systematic? Graded QC sweep, per-imaging-leg table, monopole/mask-leakage null p4p-row16ib-axis-shift · depends on p4p-row16i-full-parent-dipole | Full DESI Legacy DR8 chirality catalog (class_eq, production equivariant Z2-TTA classifier; already-committed, not re-run) row16(ii) N=20,000 injection-calibrated postprocess residual bias | imports build_projector() verbatim from pipelines/p2_chirality/generate_p4_primary_label_shuffle_strict_v1_0_257.py python3 pipelines/p4prime_chirality_test/full_parent/row16ib_qc_leg_sweep.py python3 pipelines/p4prime_chirality_test/full_parent/row16ib_figure.py | local | free (local) | ~2 min | runnable now |
Row 16 (iv-b) - chirality parity vs DESI DR1 BGS cosmic-web environment row16ivb-bgs-environment · depends on row16iv-chirality-structure | P4 chirality catalog (primary-safe, 8,474,531 rows) DESI spec-z x chirality crossmatch (P5) | python3 row16ivb_fetch_bgs.py 4 python3 row16ivb_bgs_environment.py imported | local CPU | free (local) | ~2.5 h (0.7 h download + ~1.5 h nulls) | runnable now |
Papers in this program
| Work | Code | Role | Paper |
|---|---|---|---|
| Galaxy Chirality Catalog (ApJS) | P4 | lead | view |
| Environmental Dependence of Spiral Chirality (AJ) | P5 | support | view |
External data sources
| Source | Kind | License | Link |
|---|---|---|---|
| Smith42/galaxies (HF, galaxy image dataset) | dataset | — | source ↗ |
| Galaxy Zoo 1 (GZ1) CW/CCW S3 labels | labels | — | source ↗ |
| CE-ResNet pre_desi.fits (Zenodo) | model-artifact | — | source ↗ |
| Galaxy Zoo DESI morphology predictions (Walmsley 2023) | labels | — | source ↗ |
| NOIRLab Astro Data Lab TAP (ls_dr8.tractor) | api | — | source ↗ |
| DESI DR1 (base release) | dataset | — | source ↗ |
| DESI DR1 DESIVAST value-added catalog | dataset | — | source ↗ |
| bamfai/galaxy-chirality-catalog (HF) | dataset | — | source ↗ |
| bamfai/galaxy-chirality-v2 (HF checkpoint) | model | — | source ↗ |
| bamfai/astra-desi-edr-mirror (HF) | dataset | — | source ↗ |
| P4 Zenodo archive | zenodo-archive | — | source ↗ |
Releases & DOIs
Archival records and data mirrors for every work that has one — Zenodo (permanent DOI), HuggingFace (dataset mirror), Backblaze B2 (raw-file mirror), and the GitHub source repository. A release without a minted DOI yet is cited via its pinned repository commit in the meantime, never left uncited.
| Release | Kind | Host | Link | Note |
|---|---|---|---|---|
| Galaxy chirality catalog (8.47M DESI Legacy DR8 galaxies) | Zenodo dataset | Zenodo | open ↗ | concept DOI 10.5281/zenodo.21461898 · CC-BY-4.0 |
| Galaxy chirality catalog — HuggingFace mirror | Dataset mirror | HuggingFace | open ↗ | bamfai/galaxy-chirality-catalog |
| ECH Note (torsion no-go) — archival record | Zenodo archive | Zenodo | open ↗ | concept DOI 10.5281/zenodo.21481837 · CC-BY-4.0 |
| namaster-proof — paper archival record | Zenodo archive | Zenodo | open ↗ | concept DOI 10.5281/zenodo.21481841 |
| namaster-proof — software release v0.1.7 | Software archive | Zenodo | open ↗ | commit-pinned source |
| f_NL forecast — archival record | Zenodo archive | Zenodo | open ↗ | concept DOI 10.5281/zenodo.21461880 |
| Multi-survey anomaly engine — archival record | Zenodo archive | Zenodo | open ↗ | concept DOI 10.5281/zenodo.21461887 |
| DESI DR1 anomaly-score candidate catalogue (v2, 1,244 objects) | Data release | HuggingFace | open ↗ | bamfai/bigbounce-aug-011-clean-rerun · phase3_v2/2026-09-03/ |
| DESI DR1 anomaly-score candidate catalogue (v2) — Backblaze B2 mirror | Data mirror | Backblaze B2 | open ↗ | bucket bigbounce · key prefix aug-011-clean-rerun/phase3_v2/2026-09-03/ — Zenodo DOI not yet minted, cited via pinned commit |
| NANOGrav 15-yr HD free-spectrum KDE grids (Track A3 PTA input) | External data mirror | HuggingFace | open ↗ | bamfai/bigbounce-aug-011-clean-rerun · external/nanograv15yr_kde/ · sourced from public Zenodo 10.5281/zenodo.8060824 |
| Repository source (every paper, pipeline, and manifest) | Source | GitHub | open ↗ | Hubify-Projects/bigbounce · public, MIT-adjacent per repo license |
How to run one
Every experiment row above names its own entrypoint. In general:
git clone https://github.com/Hubify-Projects/bigbounce cd bigbounce python3 -m venv .venv && source .venv/bin/activate pip install -r requirements.txt # then run the exact entrypoint from the experiment's "Scripts" cell above, e.g.: python3 research/track_a3_multichannel/pbh_abundance_fnl.py