Datasets:

raihan-js
/

sae-bilingual

License:

Dataset Preview

Split (2)

train

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

The dataset generation failed

Error code:   DatasetGenerationError
Exception:    CastError
Message:      Couldn't cast
heldout_tokens: int64
heldout_dead: int64
heldout_dead_rate: double
loss_tokens: int64
base_loss: double
sae_loss: double
loss_delta: double
secs: int64
mse: double
dead_rate: double
d_sae: int64
resampled: int64
dead: int64
tokens: int64
var_explained: double
layer: int64
l0: double
to
{'layer': Value('int64'), 'tokens': Value('int64'), 'mse': Value('float64'), 'l0': Value('float64'), 'dead': Value('int64'), 'd_sae': Value('int64'), 'dead_rate': Value('float64'), 'resampled': Value('int64'), 'var_explained': Value('float64'), 'secs': 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 1827, 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 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              heldout_tokens: int64
              heldout_dead: int64
              heldout_dead_rate: double
              loss_tokens: int64
              base_loss: double
              sae_loss: double
              loss_delta: double
              secs: int64
              mse: double
              dead_rate: double
              d_sae: int64
              resampled: int64
              dead: int64
              tokens: int64
              var_explained: double
              layer: int64
              l0: double
              to
              {'layer': Value('int64'), 'tokens': Value('int64'), 'mse': Value('float64'), 'l0': Value('float64'), 'dead': Value('int64'), 'd_sae': Value('int64'), 'dead_rate': Value('float64'), 'resampled': Value('int64'), 'var_explained': Value('float64'), 'secs': 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 1694, 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 1879, 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.

layer int64	tokens int64	mse float64	l0 float64	dead int64	d_sae int64	dead_rate float64	resampled int64	var_explained float64	secs int64
12	2,000,439	0.100296	64	15,689	32,768	0.4788	235,919	0.977	469
20	2,000,439	0.973103	64	14,267	32,768	0.4354	115,032	0.8423	500
4	2,000,439	0.009883	64	2,977	32,768	0.0909	32,348	0.9077	478

sae-bilingual

TopK sparse autoencoders (2048→32768, k=64) on llm-jp-3-1.8B residual stream (layers 4/12/20), trained streaming on Wikipedia JA+EN. For sae-bilingual.

models/sae_l*.pt: SAE weights (ours, Apache-2.0): encoder_w/bias, decoder_w, pre_bias, k. Load with the repo's src/sae/model.py.
sharing_l*.json: paired-sentence Pearson sharing on 20k Tatoeba pairs + top-100 shared features with bilingual examples.
m3*.json(l): pre-registered causal tests (ablate/steer, paired random controls) + frozen feature list. Verdict: weak JA, null EN.
browser/index.html: static feature browser (open in any browser).

Headline: sharing rises then falls with depth (r>.5 share 0.25/0.29/0.13); L4 = names, L12 = concepts, L20 = mixed incl. script features.

Downloads last month: 18

Total file size:

1.61 GB