dyt-composition-artifacts / PROVENANCE.json
lucky-verma's picture
Add artifact index for dataset viewer
f5d92a8 verified
{
"created_at_utc": "2026-05-04T03:48:31+00:00",
"files": [
{
"sanitized": false,
"sha256": "4c3c90e3a7732e2d662952081b29a55affc17da70e3e3039b7c182a1ac4c1e7f",
"size_bytes": 425,
"source_path": "CITATION.bib",
"staged_path": "CITATION.bib"
},
{
"sanitized": false,
"sha256": "81a22fc36eac6288fca513bbd5cff5483d4b03d20cb0c2105223127838c14994",
"size_bytes": 2817,
"source_path": "hf-dataset-card-draft.md",
"staged_path": "README.md"
},
{
"sanitized": false,
"sha256": "05e22a72cf990c928bc7cf35fa59741ac321b481f0e83fa14bcf05f68680bcdc",
"size_bytes": 7011,
"source_path": "generated:artifact_index",
"staged_path": "data/artifact_index.jsonl"
},
{
"sanitized": false,
"sha256": "d9b7e962cf7f075119968651e0a856ca73b7228a1a283f8b68d7d4fda3cdd232",
"size_bytes": 1618,
"source_path": "results/activation_analysis.json",
"staged_path": "results/activation_analysis.json"
},
{
"sanitized": false,
"sha256": "038699ec0852c48221aa62796c998c38928bddba245fe12386799634425bbf0e",
"size_bytes": 1014,
"source_path": "results/alpha_sweep.json",
"staged_path": "results/alpha_sweep.json"
},
{
"sanitized": false,
"sha256": "9d721c4ff939d5955e5b886235c4fd1841b3c3e6bf61b40440a5a6d7474f6942",
"size_bytes": 1740,
"source_path": "results/convergence_10k.json",
"staged_path": "results/convergence_10k.json"
},
{
"sanitized": false,
"sha256": "9d92395f143c52de2e2fa40e593f0972fd75e044148ea822e062902f1d538531",
"size_bytes": 69999,
"source_path": "results/full/activation_rank_v1/aggregate.json",
"staged_path": "results/full/activation_rank_v1/aggregate.json"
},
{
"sanitized": true,
"sha256": "7797a498acfb8f2803bdc2d062e9550f4347d33b298864c59a6f08176b31a88e",
"size_bytes": 116748,
"source_path": "results/full/all_results.json",
"staged_path": "results/full/all_results.json"
},
{
"sanitized": false,
"sha256": "4ac44ce544860ab9eb771c2defc97917064311985ed71e204751a62a285550a4",
"size_bytes": 6370,
"source_path": "results/full/crossover_predictor.json",
"staged_path": "results/full/crossover_predictor.json"
},
{
"sanitized": false,
"sha256": "974f3620ec33957cdc6afc8ebd0146b62e8874dc41136bf5b9b6c3bdd7d30ff6",
"size_bytes": 4897,
"source_path": "results/full/downstream_v5/aggregate.json",
"staged_path": "results/full/downstream_v5/aggregate.json"
},
{
"sanitized": true,
"sha256": "5881efea1ede8f29296fb7a2108f64e5a8de0f548c5c784056bff1fb46a65e93",
"size_bytes": 15815,
"source_path": "results/full/fast_extract.json",
"staged_path": "results/full/fast_extract.json"
},
{
"sanitized": false,
"sha256": "3b887dccad0b445a0bf74ff84ddc42dcac1417a19f6a117994449fd0de48f60e",
"size_bytes": 22913,
"source_path": "results/full/hessian_pyhessian_v4/aggregate.json",
"staged_path": "results/full/hessian_pyhessian_v4/aggregate.json"
},
{
"sanitized": true,
"sha256": "ce37d10cf3456d3fda311ddc97ff32f377b5272f81d9ec80e5bd6dfed37d8b61",
"size_bytes": 27980,
"source_path": "results/full/lambada_eval/aggregate.json",
"staged_path": "results/full/lambada_eval/aggregate.json"
},
{
"sanitized": true,
"sha256": "ecd533b8152d2b6e348b4f94ff487fad561cb2a552d25a5ac557223832a52689",
"size_bytes": 2281,
"source_path": "results/full/manifests/alpha_sweep_118m_summary.json",
"staged_path": "results/full/manifests/alpha_sweep_118m_summary.json"
},
{
"sanitized": false,
"sha256": "14302e82266cfe5b39cfd6c0ea755a57da256a0c41f51f027f6a6183e753ab51",
"size_bytes": 16400,
"source_path": "results/full/manifests/predictor_validation.json",
"staged_path": "results/full/manifests/predictor_validation.json"
},
{
"sanitized": false,
"sha256": "9d10175c33f9c316f97f4039cb4c9bb524459e922c017ba28c1e23ffce8990f5",
"size_bytes": 11895,
"source_path": "results/full/manifests/sig_tests.json",
"staged_path": "results/full/manifests/sig_tests.json"
},
{
"sanitized": false,
"sha256": "f5d601c707aa63e4c234ba59abc14e46baa4d4998475eb245ecfbd12e495aba7",
"size_bytes": 3127,
"source_path": "results/full/noise_stability_s3/aggregate.json",
"staged_path": "results/full/noise_stability_s3/aggregate.json"
},
{
"sanitized": false,
"sha256": "3ac8a1f87929bfea98e19cbf7ba8be4c4fca769321ae4f7024faa7a68be6cf7f",
"size_bytes": 15271,
"source_path": "results/full/noise_stability_uniform/aggregate.json",
"staged_path": "results/full/noise_stability_uniform/aggregate.json"
},
{
"sanitized": false,
"sha256": "bb871cb3e3c7cb95e8b334d042fb9cbfb5d3b0ed4bf45c6e3501cb3ef255c55a",
"size_bytes": 384041,
"source_path": "results/full/saturation_results.json",
"staged_path": "results/full/saturation_results.json"
},
{
"sanitized": false,
"sha256": "0127598467a2bb63261d2587144c2101bd58ac12330c2cff0fa3416f379d6c53",
"size_bytes": 13138,
"source_path": "results/full/scale5_metadata_check.json",
"staged_path": "results/full/scale5_metadata_check.json"
},
{
"sanitized": false,
"sha256": "ac6ee3bde5369f02c23961aa2b27e3572700526c3a5e06ba96ce55b779dfba68",
"size_bytes": 2130,
"source_path": "results/phase_diagram.json",
"staged_path": "results/phase_diagram.json"
},
{
"sanitized": false,
"sha256": "fec2c4f67bef0c77485cea5c42258952c056ba252b89b9957c49867126e48aee",
"size_bytes": 1242,
"source_path": "results/rmsnorm_results.json",
"staged_path": "results/rmsnorm_results.json"
},
{
"sanitized": false,
"sha256": "f900f06124ba36f95b3c801db285c6e7e4c8331322687067ace79b50d5437c05",
"size_bytes": 3247,
"source_path": "results/scaling.json",
"staged_path": "results/scaling.json"
},
{
"sanitized": false,
"sha256": "c46d32f450711887a3dfca4f13ab468c4e4a3d0f88b4198c608365d09cf10e4d",
"size_bytes": 2026,
"source_path": "results/train_val_gap.json",
"staged_path": "results/train_val_gap.json"
},
{
"sanitized": false,
"sha256": "af2028744c89e4af0c40b4df861cc6ebe923f2193693c15de1181fdd5f14acc4",
"size_bytes": 1975,
"source_path": "results/vit_results.json",
"staged_path": "results/vit_results.json"
}
],
"paper": {
"arxiv_id": "2604.23434",
"arxiv_url": "https://arxiv.org/abs/2604.23434",
"code_url": "https://github.com/lucky-verma/dyt-composition-study",
"doi": "10.48550/arXiv.2604.23434",
"title": "When Does Removing LayerNorm Help? Activation Bounding as a Regime-Dependent Implicit Regularizer"
},
"policy": {
"excluded": [
"raw datasets",
"model checkpoints",
"training logs",
"literature PDFs",
"private venue/process metadata",
"local paths and infrastructure identifiers",
"credential-like strings"
],
"sanitizations": [
"neutralized internal OpenWebText run labels",
"neutralized hardware/environment prose",
"removed compute-environment wording from public artifact notes"
],
"source": "public GitHub repository only",
"upload_performed": false
},
"schema_version": "1.0",
"source_repo": {
"branch": "main",
"commit": "697bb6760c34c7205295f234cb68404fa62d4668",
"dirty": false,
"remote": "https://github.com/lucky-verma/dyt-composition-study.git",
"repository": "dyt-composition-study"
}
}