Datasetpublic

federicosabbadini/patch-aliasing-bayesian

Patch-Aliasing Bayesian Analysis Data Raw and processed data from the Bayesian analysis of structural patch-aliasing in Chronos-Bolt Tiny. Companion to the model weights at federicosabbadini/chronos-bolt-patch-aliasing-models. Structure clean_15model/ # 15 preregistered (P,S) configurations full_22model/ # all 22 configurations (adds 7 robustness checks) h1_fixed_offset/ # supplementary H1 analysis with fixed offset Each run folder… See the full description on the dataset page: https://huggingface.co/datasets/federicosabbadini/patch-aliasing-bayesian.

sourceHugging Faceapache-2.0updated 1mo agoView on Hugging Face

1likes693downloads

rootREADME.md

README.md66 linesDownload Raw Back to root

1---2license: apache-2.03language:4- en5tags:6- time-series7- forecasting8- chronos9- patch-aliasing10- bayesian11size_categories:12- 10K<n<100K13---14 15# Patch-Aliasing Bayesian Analysis Data16 17Raw and processed data from the Bayesian analysis of structural patch-aliasing in Chronos-Bolt Tiny.18Companion to the model weights at [`federicosabbadini/chronos-bolt-patch-aliasing-models`](https://huggingface.co/federicosabbadini/chronos-bolt-patch-aliasing-models).19 20## Structure21 22```23clean_15model/          # 15 preregistered (P,S) configurations24full_22model/           # all 22 configurations (adds 7 robustness checks)25h1_fixed_offset/        # supplementary H1 analysis with fixed offset26```27 28Each run folder contains:29 30- `01_prior_spec.json` / `.parquet` — prior specification31- `03_recovery.parquet` — parameter recovery test32- `04_*.nc` — MCMC trace files (NetCDF/ArviZ InferenceData)33- `05_diagnostics.parquet` — convergence diagnostics34- `05_sensitivity.parquet` — prior sensitivity analysis35- `05_verdicts.parquet` — hypothesis verdicts36- `data/` — input data (contrasts, collapse profiles, MDL cells, sites)37- `data/signals/` — generated probe signals38- `figures/` — analysis figures (prior predictive, forests, combs, PPC, sensitivity)39- `run.log` — execution log40 41## Models fitted42 43| ID | Estimand | Hypothesis |44|---|---|---|45| A | Population forecast attenuation | H1 behavioural |46| B | Codelength expansion at locked frequencies | H1 representational |47| C | Phase-offset dispersion | H2 |48| D1 | Token-collapse grid labelling | H3a (stride branch) |49| D2 | Stride-scaling slope | H3a (κ_S) |50 51## Key results52 53- **H1 behavioural**: inconclusive (forecast contrast 1.08 [0.45, 2.49])54- **H1 representational**: supported on clean (codelength 1.27 [1.23, 1.31])55- **H2 phase invariance**: supported (σ_φ = 0.033, ROPE prob 0.997)56- **H3a stride branch**: supported (κ_S = 1.000 [0.992, 1.008])57- **H3b patch branch**: not identified58 59## Trace file sizes60 61The `.nc` files contain full MCMC traces (4 chains × 2000 draws). Models A and C are large (~8 GB each) due to per-observation random effects; Models B, D1, D2 are smaller (4–730 MB).62 63## License64 65Apache-2.0 — same as the Chronos models these data derive from.66