Datasets:

nmaher
/

cspo-repro-artifacts

Tags:

License:

Dataset card Data Studio Files Files and versions

xet

Community

Dataset Preview

Duplicate

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 10 new columns ({'n_events', 'n_epochs', 'TTS', 'mean_cost', 'VF', 'RP', 'iqm_return_last100', 'cost_std', 'iqm_cost_last100', 'alpha'}) and 7 missing columns ({'smoke_run_status', 'algorithm', 'note', 'registered', 'has_config', 'wall_seconds', 'repo_algo_id'}).

This happened while the csv dataset builder was generating data using

hf://datasets/nmaher/cspo-repro-artifacts/analysis/claim1_alpha_ablation.csv (at revision 9012f815f4aecc2fb2960b89abc8328629c41363), ['hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/algo_coverage.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_alpha_ablation.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_derivation_audit.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_exponent_control.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_live_probe_correction_feasible-control.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_live_probe_correction_infeasible.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_live_probe_w_feasible-control.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_live_probe_w_infeasible.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_log_consistency.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim4_per_seed.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim4_released_logs.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim4_summary.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim5_sensitivity_sweep.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim6_bootstrap.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim6_oscillation.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim6_vs_table5.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/cspo_config_vs_paper.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/env_coverage.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_equivalence.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_equivalence_control.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_equivalence_control_slopes.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_equivalence_qzero.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_equivalence_wk_sensitivity.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_rate.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_rate_fits.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_rate_summary.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/recovery_metrics_recomputed.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/sensitivity_split.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/CSPO/seed-000/progress.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/CSPO/seed-001/progress.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/CSPO/seed-002/progress.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/PPOLag/seed-000/progress.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/PPOLag/seed-001/progress.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/PPOLag/seed-002/progress.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
              alpha: double
              n_epochs: int64
              VF: double
              n_events: int64
              TTS: double
              RP: double
              cost_std: double
              mean_cost: double
              iqm_return_last100: double
              iqm_cost_last100: double
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1437
              to
              {'algorithm': Value('string'), 'repo_algo_id': Value('string'), 'registered': Value('string'), 'has_config': Value('string'), 'smoke_run_status': Value('string'), 'wall_seconds': Value('float64'), 'note': Value('string')}
              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 10 new columns ({'n_events', 'n_epochs', 'TTS', 'mean_cost', 'VF', 'RP', 'iqm_return_last100', 'cost_std', 'iqm_cost_last100', 'alpha'}) and 7 missing columns ({'smoke_run_status', 'algorithm', 'note', 'registered', 'has_config', 'wall_seconds', 'repo_algo_id'}).
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/nmaher/cspo-repro-artifacts/analysis/claim1_alpha_ablation.csv (at revision 9012f815f4aecc2fb2960b89abc8328629c41363), ['hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/algo_coverage.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_alpha_ablation.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_derivation_audit.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_exponent_control.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_live_probe_correction_feasible-control.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_live_probe_correction_infeasible.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_live_probe_w_feasible-control.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_live_probe_w_infeasible.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim1_log_consistency.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim4_per_seed.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim4_released_logs.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim4_summary.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim5_sensitivity_sweep.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim6_bootstrap.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim6_oscillation.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/claim6_vs_table5.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/cspo_config_vs_paper.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/env_coverage.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_equivalence.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_equivalence_control.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_equivalence_control_slopes.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_equivalence_qzero.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_equivalence_wk_sensitivity.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_rate.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_rate_fits.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/kkt_rate_summary.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/recovery_metrics_recomputed.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/analysis/sensitivity_split.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/CSPO/seed-000/progress.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/CSPO/seed-001/progress.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/CSPO/seed-002/progress.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/PPOLag/seed-000/progress.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/PPOLag/seed-001/progress.csv', 'hf://datasets/nmaher/cspo-repro-artifacts@9012f815f4aecc2fb2960b89abc8328629c41363/retraining_pointgoal/PPOLag/seed-002/progress.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.

algorithm string	repo_algo_id string	registered string	has_config string	smoke_run_status string	wall_seconds float64	note string
CSPO (proposed)	CSPO	Y	Y	PASS	13.5	in repo benchmark grid; method under test \| ep_ret=-1.184 ep_cost=82.5
PPO-Lag	PPOLag	Y	Y	PASS	29.1	in repo benchmark grid \| ep_ret=-0.340 ep_cost=26.5
CPPO-PID	CPPOPID	Y	Y	PASS	29.3	NOT in repo benchmark grid (run_experiment_grid.py) \| ep_ret=-0.340 ep_cost=26.5
TRPOPID	TRPOPID	Y	Y	PASS	19.5	appears only in paper Table 3, not in text baseline list; NOT in repo grid \| ep_ret=-0.218 ep_cost=53.0
CPO	CPO	Y	Y	PASS	18.5	in repo benchmark grid \| ep_ret=-0.218 ep_cost=53.0
PCPO	PCPO	Y	Y	PASS	18.9	in repo benchmark grid \| ep_ret=-0.167 ep_cost=268.0
FOCOPS	FOCOPS	Y	Y	PASS	20.3	in repo benchmark grid \| ep_ret=-0.934 ep_cost=4.5
CUP	CUP	Y	Y	PASS	22.8	in repo benchmark grid \| ep_ret=-0.743 ep_cost=18.5
P3O	P3O	Y	Y	PASS	19.5	in repo benchmark grid \| ep_ret=-0.685 ep_cost=72.5
IPO	IPO	Y	Y	PASS	21.5	NOT in repo benchmark grid \| ep_ret=-0.459 ep_cost=168.0
APPO	APPO	Y	Y	PASS	21.2	in repo benchmark grid \| ep_ret=-1.136 ep_cost=44.0
C-TRPO	CTRPO	N	N	FAIL	0	ABSENT from repo; README says run in external codebase github.com/milosen/ctrpo \| AssertionError: CTRPO doesn't exist. Please choose from ('NaturalPG', 'PolicyGradient', 'PPO', 'TRPO', 'TRPOEarlyTermina
EPO	EPO	N	N	FAIL	0	ABSENT from repo; README says run in external codebase github.com/ShiqingGao/EPOPMN \| AssertionError: EPO doesn't exist. Please choose from ('NaturalPG', 'PolicyGradient', 'PPO', 'TRPO', 'TRPOEarlyTerminate
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null
null	null	null	null	null	null	null

End of preview.

CSPO reproduction artifacts

Artifacts from an independent reproduction of CSPO: Constraint-Sensitive Policy Optimization for Safe Reinforcement Learning (ICML 2026 spotlight, OpenReview 3ySR3TCMRP).

Paper code: https://github.com/serval-uni-lu/CSPO (commit 962e696eb1f07b47ac0094bf1a2aa1299f97cc9b)
Full logbook: https://huggingface.co/spaces/nmaher/repro-cspo-constraint-sensitive-policy-optimization-for-safe-reinforcement-learning

Path	What it is
`analysis/`	All recomputed metric tables: recovery metrics (TTS/RP/VF), Claim 5 sensitivity sweep, Claim 1 derivation + live-probe audits, Claim 2 KKT equivalence + rate audits, Claim 3 environment/baseline coverage
`retraining_pointgoal/<ALGO>/seed-<N>/progress.csv`	Per-epoch OmniSafe training logs from our own CSPO vs PPO-Lag retraining on `SafetyPointGoal1-v0` (paper hyperparameters, 10M-step budget, 3 seeds, CPU-only)

All analysis scripts live in the logbook Space's Workspace tab.

Downloads last month: 30

Total file size:

2.15 MB

CSPO reproduction artifacts

Contents