Datasets:

arifmohamedkhan
/

span-extraction-adversarial-geometry

Error code:   DatasetGenerationError
Exception:    CastError
Message:      Couldn't cast
squad_sigma_0.0: struct<em_add: double, em_bia: double, n: int64, eps_add_med: double, eps_add_min: double, eps_bia_m (... 68 chars omitted)
  child 0, em_add: double
  child 1, em_bia: double
  child 2, n: int64
  child 3, eps_add_med: double
  child 4, eps_add_min: double
  child 5, eps_bia_med: double
  child 6, eps_bia_min: double
  child 7, n_add_pos: int64
  child 8, n_bia_pos: int64
squad_sigma_0.001: struct<em_add: double, em_bia: double, n: int64, eps_add_med: double, eps_add_min: double, eps_bia_m (... 68 chars omitted)
  child 0, em_add: double
  child 1, em_bia: double
  child 2, n: int64
  child 3, eps_add_med: double
  child 4, eps_add_min: double
  child 5, eps_bia_med: double
  child 6, eps_bia_min: double
  child 7, n_add_pos: int64
  child 8, n_bia_pos: int64
squad_sigma_0.005: struct<em_add: double, em_bia: double, n: int64, eps_add_med: double, eps_add_min: double, eps_bia_m (... 68 chars omitted)
  child 0, em_add: double
  child 1, em_bia: double
  child 2, n: int64
  child 3, eps_add_med: double
  child 4, eps_add_min: double
  child 5, eps_bia_med: double
  child 6, eps_bia_min: double
  child 7, n_add_pos: int64
  child 8, n_bia_pos: int64
squad_sigma_0.01: struct<em_add: double, em_bia: double, n: int64, eps_add_med: double, eps_add_min: double, eps_bia_m (... 68 chars omitted)
  child 0, em_add: double
  child 1, em_bia: double
  child 2, n: int64
  child 3, eps_add_med: double
  child 4, eps_add_min: double
  child 5, eps_bia_med: double
  child 6,
...

  child 2, n: int64
  child 3, eps_add_med: double
  child 4, eps_add_min: double
  child 5, eps_bia_med: double
  child 6, eps_bia_min: double
  child 7, n_add_pos: int64
  child 8, n_bia_pos: int64
TriviaQA_sigma_0.1: struct<em_add: double, em_bia: double, n: int64, eps_add_med: null, eps_add_min: null, eps_bia_med:  (... 60 chars omitted)
  child 0, em_add: double
  child 1, em_bia: double
  child 2, n: int64
  child 3, eps_add_med: null
  child 4, eps_add_min: null
  child 5, eps_bia_med: null
  child 6, eps_bia_min: null
  child 7, n_add_pos: int64
  child 8, n_bia_pos: int64
spectral_init: struct<em: double, eps_median: double, eps_min: double>
  child 0, em: double
  child 1, eps_median: double
  child 2, eps_min: double
trained: struct<em: double, eps_median: double, eps_min: double, n_positive: int64, final_sigma: double, fina (... 39 chars omitted)
  child 0, em: double
  child 1, eps_median: double
  child 2, eps_min: double
  child 3, n_positive: int64
  child 4, final_sigma: double
  child 5, final_u_norm: double
  child 6, final_v_norm: double
config: struct<epochs: int64, lr: double, batch_size: int64, sigma_init: double, max_train: int64>
  child 0, epochs: int64
  child 1, lr: double
  child 2, batch_size: int64
  child 3, sigma_init: double
  child 4, max_train: int64
baseline: struct<em: double, eps_median: double, eps_min: double, n_positive: int64>
  child 0, em: double
  child 1, eps_median: double
  child 2, eps_min: double
  child 3, n_positive: int64
to
{'baseline': {'em': Value('float64'), 'eps_median': Value('float64'), 'eps_min': Value('float64'), 'n_positive': Value('int64')}, 'spectral_init': {'em': Value('float64'), 'eps_median': Value('float64'), 'eps_min': Value('float64')}, 'trained': {'em': Value('float64'), 'eps_median': Value('float64'), 'eps_min': Value('float64'), 'n_positive': Value('int64'), 'final_sigma': Value('float64'), 'final_u_norm': Value('float64'), 'final_v_norm': Value('float64')}, 'config': {'epochs': Value('int64'), 'lr': Value('float64'), 'batch_size': Value('int64'), 'sigma_init': Value('float64'), 'max_train': Value('int64')}}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1816, in _prepare_split_single
                  for key, table in generator:
                                    ^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
                  for item in generator(*args, **kwargs):
                              ~~~~~~~~~^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_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
              squad_sigma_0.0: struct<em_add: double, em_bia: double, n: int64, eps_add_med: double, eps_add_min: double, eps_bia_m (... 68 chars omitted)
                child 0, em_add: double
                child 1, em_bia: double
                child 2, n: int64
                child 3, eps_add_med: double
                child 4, eps_add_min: double
                child 5, eps_bia_med: double
                child 6, eps_bia_min: double
                child 7, n_add_pos: int64
                child 8, n_bia_pos: int64
              squad_sigma_0.001: struct<em_add: double, em_bia: double, n: int64, eps_add_med: double, eps_add_min: double, eps_bia_m (... 68 chars omitted)
                child 0, em_add: double
                child 1, em_bia: double
                child 2, n: int64
                child 3, eps_add_med: double
                child 4, eps_add_min: double
                child 5, eps_bia_med: double
                child 6, eps_bia_min: double
                child 7, n_add_pos: int64
                child 8, n_bia_pos: int64
              squad_sigma_0.005: struct<em_add: double, em_bia: double, n: int64, eps_add_med: double, eps_add_min: double, eps_bia_m (... 68 chars omitted)
                child 0, em_add: double
                child 1, em_bia: double
                child 2, n: int64
                child 3, eps_add_med: double
                child 4, eps_add_min: double
                child 5, eps_bia_med: double
                child 6, eps_bia_min: double
                child 7, n_add_pos: int64
                child 8, n_bia_pos: int64
              squad_sigma_0.01: struct<em_add: double, em_bia: double, n: int64, eps_add_med: double, eps_add_min: double, eps_bia_m (... 68 chars omitted)
                child 0, em_add: double
                child 1, em_bia: double
                child 2, n: int64
                child 3, eps_add_med: double
                child 4, eps_add_min: double
                child 5, eps_bia_med: double
                child 6,
              ...
              
                child 2, n: int64
                child 3, eps_add_med: double
                child 4, eps_add_min: double
                child 5, eps_bia_med: double
                child 6, eps_bia_min: double
                child 7, n_add_pos: int64
                child 8, n_bia_pos: int64
              TriviaQA_sigma_0.1: struct<em_add: double, em_bia: double, n: int64, eps_add_med: null, eps_add_min: null, eps_bia_med:  (... 60 chars omitted)
                child 0, em_add: double
                child 1, em_bia: double
                child 2, n: int64
                child 3, eps_add_med: null
                child 4, eps_add_min: null
                child 5, eps_bia_med: null
                child 6, eps_bia_min: null
                child 7, n_add_pos: int64
                child 8, n_bia_pos: int64
              spectral_init: struct<em: double, eps_median: double, eps_min: double>
                child 0, em: double
                child 1, eps_median: double
                child 2, eps_min: double
              trained: struct<em: double, eps_median: double, eps_min: double, n_positive: int64, final_sigma: double, fina (... 39 chars omitted)
                child 0, em: double
                child 1, eps_median: double
                child 2, eps_min: double
                child 3, n_positive: int64
                child 4, final_sigma: double
                child 5, final_u_norm: double
                child 6, final_v_norm: double
              config: struct<epochs: int64, lr: double, batch_size: int64, sigma_init: double, max_train: int64>
                child 0, epochs: int64
                child 1, lr: double
                child 2, batch_size: int64
                child 3, sigma_init: double
                child 4, max_train: int64
              baseline: struct<em: double, eps_median: double, eps_min: double, n_positive: int64>
                child 0, em: double
                child 1, eps_median: double
                child 2, eps_min: double
                child 3, n_positive: int64
              to
              {'baseline': {'em': Value('float64'), 'eps_median': Value('float64'), 'eps_min': Value('float64'), 'n_positive': Value('int64')}, 'spectral_init': {'em': Value('float64'), 'eps_median': Value('float64'), 'eps_min': Value('float64')}, 'trained': {'em': Value('float64'), 'eps_median': Value('float64'), 'eps_min': Value('float64'), 'n_positive': Value('int64'), 'final_sigma': Value('float64'), 'final_u_norm': Value('float64'), 'final_v_norm': Value('float64')}, 'config': {'epochs': Value('int64'), 'lr': Value('float64'), 'batch_size': Value('int64'), 'sigma_init': Value('float64'), 'max_train': Value('int64')}}
              because column names don't match
              
              The above exception was the direct cause of the following exception:
              
              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 1869, in _prepare_split_single
                  raise DatasetGenerationError("An error occurred while generating the dataset") from e
              datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset

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.

baseline dict	spectral_init dict	trained dict	config dict
{ "em": 0.3, "eps_median": 1.8271093368530273, "eps_min": 0.00003250884037697688, "n_positive": 500 }	{ "em": 0.3, "eps_median": 1.8115628957748413, "eps_min": 7.803024573149742e-7 }	{ "em": 0.292, "eps_median": 0.7736152410507202, "eps_min": 0.000015201909263851121, "n_positive": 500, "final_sigma": -0.004876310471445322, "final_u_norm": 2.0297563076019287, "final_v_norm": 1.8992451429367065 }	{ "epochs": 3, "lr": 0.00005, "batch_size": 4, "sigma_init": 0.001, "max_train": 5000 }

Span Extraction Adversarial Geometry

Reproducibility data for "Why Additive Span Extraction Heads Cannot Be Improved by Coupling: Exact Adversarial Radii, Sign Attacks, and Encoder-Level Certified Training."

Key Results

Method	EM	ε* median	Δε*
Additive baseline	59.8%	2.17	—
Elsayed (global margin)	65.6%	2.96	+36%
CBCT (RoBERTa)	66.0%	3.31	+52%
CBCT (BERT-large)	62.0%	4.71	+113%
CBCT (DistilBERT)	57.8%	2.96	+45%
Biaffine spectral	65.4%	0.93	-57%

Statistical tests: CBCT vs baseline Wilcoxon p = 10⁻⁴², CBCT vs Elsayed p = 10⁻⁶.

data/logits/: 14,863 examples × 10 extraction heads × 6 configs
data/hidden_states/: Hidden state vectors for SQuAD examples
data/results/: All experiment outputs (baseline comparison, multi-model CBCT, paired tests, theory verification, hyperparameter sensitivity, bucket analysis, cross-dataset)
verify_all.py: Checks every number in the paper against data files

Full code release on GitHub after acceptance.

Verification

python verify_all.py

Citation

@article{ahamad2026span,
  title={Why Additive Span Extraction Heads Cannot Be Improved by Coupling:
         Exact Adversarial Radii, Sign Attacks, and Encoder-Level Certified Training},
  author={Ahamad, Arif Mohamed Khan Rabi and Douangnouanexay, Laddaphone},
  year={2026}
}

Downloads last month: 89

Total file size:

234 MB

Span Extraction Adversarial Geometry

Key Results

Contents

Verification

Citation