Dataset Preview

Split (1)

train

The full dataset viewer is not available (click to read why). Only showing a preview of the rows.

The dataset generation failed because of a cast error

Error code:   DatasetGenerationCastError
Exception:    DatasetGenerationCastError
Message:      An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 4 new columns ({'panel', 'sim', 'theory', 's'}) and 5 missing columns ({'theory_R', 'n', 'sim_risk', 'theory_B', 'theory_V'}).

This happened while the csv dataset builder was generating data using

hf://datasets/Eishaan/selfdistill-repro-bundle/outputs/claim2/claim2_all.csv (at revision bd8608ad192dcdd13b52e97545a2086d3fbffd67), ['hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim1/claim1_BV_decomposition.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim1/claim1_risk_theory_vs_sim.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim2/claim2_all.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim2/claim2_fig1b_ushape.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim2/claim2_fig3a_anisotropy.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim3/claim3_multispike_selection.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim3/claim3_noise_suppression.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim3/claim3_survival_factor.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim4/claim4_igcv_vs_risk.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim4/claim4_selection.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim5/claim5_cifar.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/general/general_fig5b_ushape.csv']

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1837, in _prepare_split_single
                  writer.write_table(table)
                  ~~~~~~~~~~~~~~~~~~^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
                  self._write_table(pa_table, writer_batch_size=writer_batch_size)
                  ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
                  pa_table = table_cast(pa_table, self._schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              panel: string
              s: double
              rho: double
              t: int64
              sim: double
              sim_std: double
              theory: double
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1032
              to
              {'rho': Value('float64'), 'n': Value('int64'), 't': Value('int64'), 'sim_risk': Value('float64'), 'sim_std': Value('float64'), 'theory_R': Value('float64'), 'theory_B': Value('float64'), 'theory_V': Value('float64')}
              because column names don't match
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1683, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1839, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
                  ...<4 lines>...
                  )
              datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
              
              All the data files must have the same columns, but at some point there are 4 new columns ({'panel', 'sim', 'theory', 's'}) and 5 missing columns ({'theory_R', 'n', 'sim_risk', 'theory_B', 'theory_V'}).
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/Eishaan/selfdistill-repro-bundle/outputs/claim2/claim2_all.csv (at revision bd8608ad192dcdd13b52e97545a2086d3fbffd67), ['hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim1/claim1_BV_decomposition.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim1/claim1_risk_theory_vs_sim.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim2/claim2_all.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim2/claim2_fig1b_ushape.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim2/claim2_fig3a_anisotropy.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim3/claim3_multispike_selection.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim3/claim3_noise_suppression.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim3/claim3_survival_factor.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim4/claim4_igcv_vs_risk.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim4/claim4_selection.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/claim5/claim5_cifar.csv', 'hf://datasets/Eishaan/selfdistill-repro-bundle@bd8608ad192dcdd13b52e97545a2086d3fbffd67/outputs/general/general_fig5b_ushape.csv']
              
              Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

rho float64	n int64	t int64	sim_risk float64	sim_std float64	theory_R float64	theory_B float64	theory_V float64
1.5	800	0	2.115912	0.080559	2.028835	0.009612	2.019223
1.5	800	1	1.459677	0.081771	1.402322	0.037697	1.364626
1.5	800	2	1.054863	0.05565	1.010678	0.083168	0.92751
1.5	800	3	0.828184	0.054825	0.780398	0.144988	0.63541
1.5	800	4	0.684399	0.043781	0.66218	0.222167	0.440013
1.5	800	5	0.641435	0.045741	0.622872	0.31376	0.309112
1.5	800	6	0.663317	0.053661	0.640099	0.418867	0.221232
1.5	800	7	0.721703	0.066246	0.698684	0.536627	0.162057
1.5	800	8	0.796689	0.067166	0.788262	0.666221	0.122041
1.5	800	9	0.908851	0.072028	0.901689	0.806868	0.09482
1.5	800	10	1.026127	0.072698	1.033974	0.957824	0.07615
1.5	800	11	1.178952	0.081056	1.18158	1.118378	0.063201
1.5	800	12	1.342181	0.085586	1.341942	1.287856	0.054086
1.5	800	13	1.513978	0.077431	1.513157	1.465613	0.047545
1.5	800	14	1.694504	0.086568	1.693774	1.651036	0.042738
2	600	0	1.081921	0.116302	1.073964	0.036982	1.036982
2	600	1	0.699712	0.072485	0.694977	0.142294	0.552683
2	600	2	0.625614	0.077927	0.615999	0.308045	0.307954
2	600	3	0.725062	0.113003	0.71025	0.527046	0.183205
2	600	4	0.917726	0.143519	0.911374	0.792749	0.118625
2	600	5	1.165143	0.160004	1.183496	1.0992	0.084296
2	600	6	1.488715	0.177318	1.506234	1.440986	0.065247
2	600	7	1.837833	0.184195	1.867174	1.813194	0.05398
2	600	8	2.228579	0.198502	2.258101	2.211366	0.046735
2	600	9	2.641269	0.208731	2.673088	2.631465	0.041623
2	600	10	3.065084	0.255521	3.107528	3.069838	0.03769
2	600	11	3.51613	0.267085	3.557634	3.523184	0.03445
2	600	12	3.980691	0.286294	4.020179	3.988527	0.031652
2	600	13	4.458014	0.318097	4.492348	4.463183	0.029165
2	600	14	4.938186	0.288802	4.971657	4.944742	0.026915
2.5	480	0	0.818694	0.0445	0.800166	0.0801	0.720066
2.5	480	1	0.639394	0.055164	0.638072	0.30252	0.335552
2.5	480	2	0.828759	0.069717	0.819568	0.643049	0.176519
2.5	480	3	1.173552	0.128301	1.188878	1.080626	0.108252
2.5	480	4	1.65916	0.187117	1.673765	1.596961	0.076805
2.5	480	5	2.226063	0.187863	2.236747	2.176207	0.06054
2.5	480	6	2.82833	0.171536	2.855422	2.804668	0.050754
2.5	480	7	3.512823	0.183971	3.514459	3.470539	0.04392
2.5	480	8	4.214571	0.152404	4.20226	4.163672	0.038589
2.5	480	9	4.903872	0.141557	4.909523	4.875379	0.034144
2.5	480	10	5.634946	0.166183	5.628559	5.598251	0.030308
2.5	480	11	6.358171	0.194144	6.352937	6.325996	0.026942
2.5	480	12	7.109372	0.187436	7.077265	7.0533	0.023965
2.5	480	13	7.809839	0.194076	7.797029	7.775705	0.021324
2.5	480	14	8.543225	0.236624	8.508476	8.4895	0.018976
1.5	800	0	2.115912	0.080559	2.028835	0.009612	2.019223
1.5	800	1	1.459677	0.081771	1.402322	0.037697	1.364626
1.5	800	2	1.054863	0.05565	1.010678	0.083168	0.92751
1.5	800	3	0.828184	0.054825	0.780398	0.144988	0.63541
1.5	800	4	0.684399	0.043781	0.66218	0.222167	0.440013
1.5	800	5	0.641435	0.045741	0.622872	0.31376	0.309112
1.5	800	6	0.663317	0.053661	0.640099	0.418867	0.221232
1.5	800	7	0.721703	0.066246	0.698684	0.536627	0.162057
1.5	800	8	0.796689	0.067166	0.788262	0.666221	0.122041
1.5	800	9	0.908851	0.072028	0.901689	0.806868	0.09482
1.5	800	10	1.026127	0.072698	1.033974	0.957824	0.07615
1.5	800	11	1.178952	0.081056	1.18158	1.118378	0.063201
1.5	800	12	1.342181	0.085586	1.341942	1.287856	0.054086
1.5	800	13	1.513978	0.077431	1.513157	1.465613	0.047545
1.5	800	14	1.694504	0.086568	1.693774	1.651036	0.042738
2	600	0	1.081921	0.116302	1.073964	0.036982	1.036982
2	600	1	0.699712	0.072485	0.694977	0.142294	0.552683
2	600	2	0.625614	0.077927	0.615999	0.308045	0.307954
2	600	3	0.725062	0.113003	0.71025	0.527046	0.183205
2	600	4	0.917726	0.143519	0.911374	0.792749	0.118625
2	600	5	1.165143	0.160004	1.183496	1.0992	0.084296
2	600	6	1.488715	0.177318	1.506234	1.440986	0.065247
2	600	7	1.837833	0.184195	1.867174	1.813194	0.05398
2	600	8	2.228579	0.198502	2.258101	2.211366	0.046735
2	600	9	2.641269	0.208731	2.673088	2.631465	0.041623
2	600	10	3.065084	0.255521	3.107528	3.069838	0.03769
2	600	11	3.51613	0.267085	3.557634	3.523184	0.03445
2	600	12	3.980691	0.286294	4.020179	3.988527	0.031652
2	600	13	4.458014	0.318097	4.492348	4.463183	0.029165
2	600	14	4.938186	0.288802	4.971657	4.944742	0.026915
2.5	480	0	0.818694	0.0445	0.800166	0.0801	0.720066
2.5	480	1	0.639394	0.055164	0.638072	0.30252	0.335552
2.5	480	2	0.828759	0.069717	0.819568	0.643049	0.176519
2.5	480	3	1.173552	0.128301	1.188878	1.080626	0.108252
2.5	480	4	1.65916	0.187117	1.673765	1.596961	0.076805
2.5	480	5	2.226063	0.187863	2.236747	2.176207	0.06054
2.5	480	6	2.82833	0.171536	2.855422	2.804668	0.050754
2.5	480	7	3.512823	0.183971	3.514459	3.470539	0.04392
2.5	480	8	4.214571	0.152404	4.20226	4.163672	0.038589
2.5	480	9	4.903872	0.141557	4.909523	4.875379	0.034144
2.5	480	10	5.634946	0.166183	5.628559	5.598251	0.030308
2.5	480	11	6.358171	0.194144	6.352937	6.325996	0.026942
2.5	480	12	7.109372	0.187436	7.077265	7.0533	0.023965
2.5	480	13	7.809839	0.194076	7.797029	7.775705	0.021324
2.5	480	14	8.543225	0.236624	8.508476	8.4895	0.018976
1.5	null	0	null	0.107975	null	null	null
1.5	null	1	null	0.092613	null	null	null
1.5	null	2	null	0.085458	null	null	null
1.5	null	3	null	0.090845	null	null	null
1.5	null	4	null	0.090522	null	null	null
1.5	null	5	null	0.089775	null	null	null
1.5	null	6	null	0.109498	null	null	null
1.5	null	7	null	0.120517	null	null	null
1.5	null	8	null	0.124283	null	null	null
1.5	null	9	null	0.123617	null	null	null

End of preview.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Reproduction — Why Self-Training Helps and Hurts: Denoising vs. Signal Forgetting

Independent reproduction of ICML 2026 paper VnA5q5jXVz (arXiv 2602.14029), Wu, Yang & Sun. No official code was released; everything here is written from the paper.

What is verified

Claim	Content	Status	Evidence
1	Deterministic-equivalent recursion `R_t = B_t (forgetting, up) + V*_t (noise, down exp)`	verified	`claim1_decomposition.py`, `general_cov.py` (spiked Thm 3.2 and general Thm 4.2)
2	U-shaped risk & optimal early stopping, strictly under anisotropy (s>1)	verified	`claim2_ushape.py` (Fig 1b, Fig 3a)
3	Direction-dependent spectral filter: survival `(s/(s+tau))^{t+1}`, noise `(1+tau)^{-t}`	verified	`claim3_spectral.py`
4	iGCV consistently estimates risk & recovers t* with no validation set	verified	`claim4_igcv.py`
5	Trade-off persists for deep nets (ResNet-50 / CIFAR-10 self-distillation)	verified (GPU Job)	`cifar_selfdistill.py` (Fig 6a)

Layout

src/linear_selftrain.py — Algorithm 1 (ridgeless/ridge self-training), spiked Thm 3.2, multi-spike Thm 3.6, general deterministic-equivalent recursion (Def 4.1 / eq 10), iGCV (eq 11-12), fast structured samplers.
src/plotting.py — Plotly + CSV export helpers.
claim{1,2,3,4}_*.py, general_cov.py — per-claim reproductions (linear theory, CPU).
cifar_selfdistill.py — PEP-723 UV script for the ResNet-50/CIFAR-10 experiment (HF GPU Job).
outputs/ — generated figures (HTML) + raw data (CSV) per claim.

Reproduce

# linear-theory claims (CPU, ~1 min each). WSL2 note: single-thread BLAS is faster.
uv run --env-file .env python claim1_decomposition.py   # .env pins OPENBLAS_NUM_THREADS=1
uv run --env-file .env python claim2_ushape.py
uv run --env-file .env python claim3_spectral.py
uv run --env-file .env python claim4_igcv.py
uv run --env-file .env python general_cov.py

# deep-net claim (GPU): run on Hugging Face Jobs
hf jobs uv run --flavor a10g-small --secrets HF_TOKEN cifar_selfdistill.py \
  --n 10000 --K 4 --epochs 40 --etas 0.4,0.6,0.8

Downloads last month: 214

Total file size:

2.78 MB

Collection including Eishaan/selfdistill-repro-bundle

ICML 2026 Repro - Why Self-Training Helps and Hurts

Collection

Reproduction artifacts for ICML 2026 paper VnA5q5jXVz (arXiv 2602.14029): linear-regression theory + ResNet-50/CIFAR-10 self-distillation. • 3 items • Updated 25 days ago