Hugging Face's logo

Datasets:
vimarsh
/
icml26332-fullbatch-results

ArXiv:
Tags:
icml2026-repro
single-index-models
gradient-descent
theory-reproduction
License:
Dataset card Data Studio Files
xet
 Community
Dataset Preview
Duplicate
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 3 new columns ({'mean', 'sem', 'delta'}) and 5 missing columns ({'sq_overlap', 'step', 'dist2', 'norm', 'loss'}).

This happened while the csv dataset builder was generating data using

hf://datasets/vimarsh/icml26332-fullbatch-results/agg_online_quad.csv (at revision 5e21ee9a06603d997b54dbdeb5206694e101a513), ['hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_gd_trunc_r15.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_gd_trunc_r2.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_online_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_online_trunc.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_trunc.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/audit_eq313.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/audit_indicator.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/audit_spectrum.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/audit_uniform_bbp.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_control_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_eta_scaling.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_lr_probe.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_phases.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_phases_r15.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_quad_r2_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_quad_r2_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_time_thresholds.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_time_thresholds_r15.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.025_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.025_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.05_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.05_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.2_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.2_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.64_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.64_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.8_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.8_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_r15_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_r15_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_r2_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_r2_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_1.5_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_1.5_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_10_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_10_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_1_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_1_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_2_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_2_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_3_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_3_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_4_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_4_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_6_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_6_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_1.5_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_1.5_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_10_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_10_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_1_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_1_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_2_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_2_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_3_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_3_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_4_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_4_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_6_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_6_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/scaling_collapse.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/smooth_vs_hard.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/success_frac_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/success_frac_trunc.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/sweep_online_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/sweep_online_trunc.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/sweep_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/sweep_smooth.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/sweep_trunc.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/thm32_bound.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/threshold_fits.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/thresholds.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
              d: int64
              delta: double
              mean: double
              sem: double
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 705
              to
              {'d': Value('int64'), 'step': Value('int64'), 'sq_overlap': Value('float64'), 'norm': Value('float64'), 'dist2': Value('float64'), 'loss': 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 3 new columns ({'mean', 'sem', 'delta'}) and 5 missing columns ({'sq_overlap', 'step', 'dist2', 'norm', 'loss'}).
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/vimarsh/icml26332-fullbatch-results/agg_online_quad.csv (at revision 5e21ee9a06603d997b54dbdeb5206694e101a513), ['hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_gd_trunc_r15.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_gd_trunc_r2.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_online_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_online_trunc.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/agg_trunc.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/audit_eq313.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/audit_indicator.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/audit_spectrum.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/audit_uniform_bbp.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_control_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_eta_scaling.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_lr_probe.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_phases.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_phases_r15.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_quad_r2_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_quad_r2_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_time_thresholds.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_time_thresholds_r15.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.025_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.025_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.05_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.05_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.2_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.2_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.64_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.64_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.8_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_eta0.8_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_r15_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_r15_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_r2_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gd_trunc_r2_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_1.5_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_1.5_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_10_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_10_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_1_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_1_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_2_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_2_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_3_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_3_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_4_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_4_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_6_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_quad_6_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_1.5_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_1.5_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_10_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_10_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_1_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_1_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_2_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_2_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_3_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_3_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_4_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_4_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_6_summary.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/gdd_trunc_6_traj.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/scaling_collapse.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/smooth_vs_hard.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/success_frac_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/success_frac_trunc.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/sweep_online_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/sweep_online_trunc.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/sweep_quad.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/sweep_smooth.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/sweep_trunc.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/thm32_bound.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/threshold_fits.csv', 'hf://datasets/vimarsh/icml26332-fullbatch-results@5e21ee9a06603d997b54dbdeb5206694e101a513/thresholds.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.

d
int64
step
int64
sq_overlap
float64
norm
float64
dist2
float64
loss
float64
64
0
0.009831
0
1
1.389713
64
10
0.011109
0
1
1.389713
64
20
0.012597
0
1
1.389713
64
30
0.014327
0
1
1.389713
64
40
0.01634
0
1
1.389713
64
50
0.01868
0
1
1.389713
64
60
0.021398
0
1
1.389713
64
70
0.02455
0
1
1.389713
64
80
0.028198
0
1
1.389713
64
90
0.03241
0
1
1.389713
64
100
0.03726
0
1
1.389713
64
110
0.042823
0
1
1.389713
64
120
0.04918
0
1
1.389713
64
130
0.056409
0
1
1.389713
64
140
0.064584
0
1
1.389713
64
150
0.073772
0
1
1.389713
64
160
0.084027
0
1
1.389713
64
170
0.095386
0
1
1.389713
64
180
0.107863
0
1
1.389713
64
190
0.121442
0
1
1.389713
64
200
0.13608
0
1
1.389713
64
210
0.151698
0
1
1.389713
64
220
0.168186
0
1
1.389713
64
230
0.185406
0
1
1.389713
64
240
0.203195
0
1
1.389713
64
250
0.221375
0
1
1.389713
64
260
0.239763
0
1
1.389713
64
270
0.258175
0
1
1.389713
64
280
0.276438
0
1
1.389713
64
290
0.294397
0
1
1.389713
64
300
0.311916
0
1
1.389713
64
310
0.328883
0
1
1.389713
64
320
0.345209
0
1
1.389713
64
330
0.36083
0
1
1.389713
64
340
0.375701
0
1
1.389713
64
350
0.389799
0
1
1.389713
64
360
0.403113
0
1
1.389713
64
370
0.415647
0
1
1.389713
64
380
0.427416
0
1
1.389713
64
390
0.438441
0
1
1.389713
64
400
0.448751
0
1
1.389713
64
410
0.458379
0
1
1.389713
64
420
0.467359
0
1
1.389713
64
430
0.475728
0
1
1.389713
64
440
0.483526
0
1
1.389713
64
450
0.490789
0
1
1.389713
64
460
0.497555
0
1
1.389713
64
470
0.503862
0
1
1.389713
64
480
0.509744
0
1
1.389713
64
490
0.515235
0
1
1.389713
64
500
0.520368
0
1
1.389713
64
510
0.525174
0
1
1.389713
64
520
0.529681
0
1
1.389713
64
530
0.533916
0
1
1.389713
64
540
0.537906
0
1
1.389713
64
550
0.541673
0
1
1.389713
64
560
0.54524
0
1
1.389713
64
570
0.548626
0
1
1.389713
64
580
0.55185
0
1
1.389713
64
590
0.554931
0
1
1.389713
64
600
0.557882
0
1
1.389713
64
610
0.56072
0
1
1.389713
64
620
0.563457
0
1
1.389713
64
630
0.566105
0
1
1.389713
64
640
0.568676
0
1
1.389713
64
650
0.571178
0
1
1.389713
64
660
0.573622
0
1
1.389713
64
670
0.576015
0
1
1.389713
64
680
0.578364
0
1
1.389713
64
690
0.580675
0
1
1.389713
64
700
0.582956
0
1
1.389713
64
710
0.585209
0
1
1.389713
64
720
0.587441
0
1
1.389713
64
730
0.589654
0
1
1.389713
64
740
0.591852
0
1
1.389713
64
750
0.594038
0
1
1.389713
64
760
0.596213
0
1
1.389713
64
770
0.598381
0
1
1.389713
64
780
0.600542
0
1
1.389713
64
790
0.602697
0
1
1.389713
64
800
0.604847
0
1
1.389713
64
810
0.606993
0
1
1.389713
64
820
0.609134
0
1
1.389713
64
830
0.611272
0
1
1.389713
64
840
0.613404
0
1
1.389713
64
850
0.615531
0
1
1.389713
64
860
0.617653
0
1
1.389713
64
870
0.619768
0
1
1.389713
64
880
0.621874
0
1
1.389713
64
890
0.623973
0
1
1.389713
64
900
0.626061
0
1
1.389713
64
910
0.628137
0
1
1.389713
64
920
0.630201
0
1
1.389713
64
930
0.632251
0
1
1.389713
64
940
0.634286
0
1
1.389713
64
950
0.636303
0
1
1.389713
64
960
0.638303
0
1
1.389713
64
970
0.640283
0
1
1.389713
64
980
0.642241
0
1
1.389713
64
990
0.644178
0
1
1.389713
End of preview.

Reproduction results — "Full-Batch Gradient Descent Outperforms One-Pass SGD" (ICML 2026 #26332)

Raw and aggregated results for an independent reproduction of arXiv:2602.02431 / OpenReview QItZDBVCT0.

All numbers were produced locally on 2x NVIDIA RTX 4000 Ada with the scripts in the reproduction logbook (scripts/sim.py, scripts/sweep_spherical.py, scripts/sweep_online_sgd.py, scripts/sweep_squared_gd.py, scripts/spectral_audit.py, scripts/thm32_bound.py).

file contents
sweep_quad.csv full-batch spherical GD, σ(z)=z², per-run squared overlap (Claim 1)
sweep_trunc.csv full-batch spherical GD, σ(z)=min(z²,8) (Claims 2/5)
sweep_smooth.csv same with the C^∞ truncation of paper eq. (3.10) (robustness)
sweep_online_{trunc,quad}.csv one-pass spherical SGD baseline over a grid of η=c/d (Claim 5)
gd_trunc_r2_{traj,summary}.csv squared-loss full-batch GD, r₀=d⁻² (Claims 3/4)
gd_trunc_r15_{traj,summary}.csv squared-loss full-batch GD, r₀=d⁻¹⁵ (exact Theorem 4.1 setting)
gd_quad_r2_*, gdd_* untruncated-activation and δ-sweep controls
gd_trunc_eta*_* learning-rate sweep for the O(log d / η) phase-1 length
audit_{spectrum,uniform_bbp,indicator}.csv numerical audits of the spectral statements
thm32_bound.csv audit of the Theorem 3.2 deficit bound 1 − C(e^{−M/2} + (d/n)^{1/5})
agg_*.csv, thresholds.csv, threshold_fits.csv aggregated curves and log-d fits

Logbook: https://huggingface.co/spaces/vimarsh/repro-full-batch-gd-outperforms-one-pass-sgd-sample-complexity-separation
Downloads last month
69
Total file size:
90.9 MB

Paper for vimarsh/icml26332-fullbatch-results