The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
schema: string
kind: string
started_utc: string
base: string
base_revision: string
adapter_repo: string
adapter_dir: string
dataset: string
max_length: int64
seed: int64
method: string
gpu: string
pins: struct<eval_sha256: string, train_sha256: string, eval_sha256_expected: string, train_sha256_expecte (... 10 chars omitted)
child 0, eval_sha256: string
child 1, train_sha256: string
child 2, eval_sha256_expected: string
child 3, train_sha256_expected: string
adapter_repo_revision: string
rows: struct<eval: int64, train_ref: int64, train_ref_note: string>
child 0, eval: int64
child 1, train_ref: int64
child 2, train_ref_note: string
eval_base: struct<rows: int64, tokens: int64, mean_loss: double, perplexity: double, token_accuracy: double, se (... 53 chars omitted)
child 0, rows: int64
child 1, tokens: int64
child 2, mean_loss: double
child 3, perplexity: double
child 4, token_accuracy: double
child 5, seconds: double
child 6, energy_j: int64
child 7, energy_label: string
eval_adapter: struct<rows: int64, tokens: int64, mean_loss: double, perplexity: double, token_accuracy: double, se (... 53 chars omitted)
child 0, rows: int64
child 1, tokens: int64
child 2, mean_loss: double
child 3, perplexity: double
child 4, token_accuracy: double
child 5, seconds: double
child 6, energy_j: int64
child 7, energy_label: string
train_ref_adapter: struct<rows: int64, tokens: int64, mean_loss: double, perplexity: double, token_accuracy: double, se
...
ane: string
heldout: struct<verdicts: struct<criterion_2_contamination: struct<pass: bool, rate: double, rows_used: int64 (... 379 chars omitted)
child 0, verdicts: struct<criterion_2_contamination: struct<pass: bool, rate: double, rows_used: int64, rule: string>, (... 361 chars omitted)
child 0, criterion_2_contamination: struct<pass: bool, rate: double, rows_used: int64, rule: string>
child 0, pass: bool
child 1, rate: double
child 2, rows_used: int64
child 3, rule: string
child 1, criterion_3_refusal: struct<adapter_refusal_rate: double, base_refusal_rate: double, pass: bool, rule: string>
child 0, adapter_refusal_rate: double
child 1, base_refusal_rate: double
child 2, pass: bool
child 3, rule: string
child 2, criterion_4_neutral: struct<adapter_accuracy: double, base_accuracy: double, pass: bool, rule: string>
child 0, adapter_accuracy: double
child 1, base_accuracy: double
child 2, pass: bool
child 3, rule: string
child 3, criterion_5_invented_id: struct<invented_id_rate: double, pass: bool, rule: string, tau: double>
child 0, invented_id_rate: double
child 1, pass: bool
child 2, rule: string
child 3, tau: double
child 4, heldout_quality_all_pass: bool
child 5, status: string
bench_receipt_sha256: string
generated_utc: timestamp[s]
generator: string
train_report_sha256: null
to
{'base': Value('string'), 'bench_receipt_sha256': Value('string'), 'criterion_1': {'eval_loss': Value('float64'), 'pass': Value('bool'), 'rule': Value('string'), 'train_ref_loss': Value('float64')}, 'dataset': Value('string'), 'eval_adapter': {'energy_j': Value('int64'), 'energy_label': Value('string'), 'mean_loss': Value('float64'), 'perplexity': Value('float64'), 'rows': Value('int64'), 'seconds': Value('float64'), 'token_accuracy': Value('float64'), 'tokens': Value('int64')}, 'eval_base': {'energy_j': Value('int64'), 'energy_label': Value('string'), 'mean_loss': Value('float64'), 'perplexity': Value('float64'), 'rows': Value('int64'), 'seconds': Value('float64'), 'token_accuracy': Value('float64'), 'tokens': Value('int64')}, 'generated_utc': Value('timestamp[s]'), 'generator': Value('string'), 'heldout': {'verdicts': {'criterion_2_contamination': {'pass': Value('bool'), 'rate': Value('float64'), 'rows_used': Value('int64'), 'rule': Value('string')}, 'criterion_3_refusal': {'adapter_refusal_rate': Value('float64'), 'base_refusal_rate': Value('float64'), 'pass': Value('bool'), 'rule': Value('string')}, 'criterion_4_neutral': {'adapter_accuracy': Value('float64'), 'base_accuracy': Value('float64'), 'pass': Value('bool'), 'rule': Value('string')}, 'criterion_5_invented_id': {'invented_id_rate': Value('float64'), 'pass': Value('bool'), 'rule': Value('string'), 'tau': Value('float64')}, 'heldout_quality_all_pass': Value('bool'), 'status': Value('string')}}, 'heldout_receipt_sha256': Value('string'), 'improvement_vs_base': {'delta': Value('float64'), 'eval_loss_adapter': Value('float64'), 'eval_loss_base': Value('float64')}, 'lane': Value('string'), 'schema': Value('string'), 'train_ref_adapter': {'energy_j': Value('int64'), 'energy_label': Value('string'), 'mean_loss': Value('float64'), 'perplexity': Value('float64'), 'rows': Value('int64'), 'seconds': Value('float64'), 'token_accuracy': Value('float64'), 'tokens': Value('int64')}, 'train_report_sha256': Value('null')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, 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 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
schema: string
kind: string
started_utc: string
base: string
base_revision: string
adapter_repo: string
adapter_dir: string
dataset: string
max_length: int64
seed: int64
method: string
gpu: string
pins: struct<eval_sha256: string, train_sha256: string, eval_sha256_expected: string, train_sha256_expecte (... 10 chars omitted)
child 0, eval_sha256: string
child 1, train_sha256: string
child 2, eval_sha256_expected: string
child 3, train_sha256_expected: string
adapter_repo_revision: string
rows: struct<eval: int64, train_ref: int64, train_ref_note: string>
child 0, eval: int64
child 1, train_ref: int64
child 2, train_ref_note: string
eval_base: struct<rows: int64, tokens: int64, mean_loss: double, perplexity: double, token_accuracy: double, se (... 53 chars omitted)
child 0, rows: int64
child 1, tokens: int64
child 2, mean_loss: double
child 3, perplexity: double
child 4, token_accuracy: double
child 5, seconds: double
child 6, energy_j: int64
child 7, energy_label: string
eval_adapter: struct<rows: int64, tokens: int64, mean_loss: double, perplexity: double, token_accuracy: double, se (... 53 chars omitted)
child 0, rows: int64
child 1, tokens: int64
child 2, mean_loss: double
child 3, perplexity: double
child 4, token_accuracy: double
child 5, seconds: double
child 6, energy_j: int64
child 7, energy_label: string
train_ref_adapter: struct<rows: int64, tokens: int64, mean_loss: double, perplexity: double, token_accuracy: double, se
...
ane: string
heldout: struct<verdicts: struct<criterion_2_contamination: struct<pass: bool, rate: double, rows_used: int64 (... 379 chars omitted)
child 0, verdicts: struct<criterion_2_contamination: struct<pass: bool, rate: double, rows_used: int64, rule: string>, (... 361 chars omitted)
child 0, criterion_2_contamination: struct<pass: bool, rate: double, rows_used: int64, rule: string>
child 0, pass: bool
child 1, rate: double
child 2, rows_used: int64
child 3, rule: string
child 1, criterion_3_refusal: struct<adapter_refusal_rate: double, base_refusal_rate: double, pass: bool, rule: string>
child 0, adapter_refusal_rate: double
child 1, base_refusal_rate: double
child 2, pass: bool
child 3, rule: string
child 2, criterion_4_neutral: struct<adapter_accuracy: double, base_accuracy: double, pass: bool, rule: string>
child 0, adapter_accuracy: double
child 1, base_accuracy: double
child 2, pass: bool
child 3, rule: string
child 3, criterion_5_invented_id: struct<invented_id_rate: double, pass: bool, rule: string, tau: double>
child 0, invented_id_rate: double
child 1, pass: bool
child 2, rule: string
child 3, tau: double
child 4, heldout_quality_all_pass: bool
child 5, status: string
bench_receipt_sha256: string
generated_utc: timestamp[s]
generator: string
train_report_sha256: null
to
{'base': Value('string'), 'bench_receipt_sha256': Value('string'), 'criterion_1': {'eval_loss': Value('float64'), 'pass': Value('bool'), 'rule': Value('string'), 'train_ref_loss': Value('float64')}, 'dataset': Value('string'), 'eval_adapter': {'energy_j': Value('int64'), 'energy_label': Value('string'), 'mean_loss': Value('float64'), 'perplexity': Value('float64'), 'rows': Value('int64'), 'seconds': Value('float64'), 'token_accuracy': Value('float64'), 'tokens': Value('int64')}, 'eval_base': {'energy_j': Value('int64'), 'energy_label': Value('string'), 'mean_loss': Value('float64'), 'perplexity': Value('float64'), 'rows': Value('int64'), 'seconds': Value('float64'), 'token_accuracy': Value('float64'), 'tokens': Value('int64')}, 'generated_utc': Value('timestamp[s]'), 'generator': Value('string'), 'heldout': {'verdicts': {'criterion_2_contamination': {'pass': Value('bool'), 'rate': Value('float64'), 'rows_used': Value('int64'), 'rule': Value('string')}, 'criterion_3_refusal': {'adapter_refusal_rate': Value('float64'), 'base_refusal_rate': Value('float64'), 'pass': Value('bool'), 'rule': Value('string')}, 'criterion_4_neutral': {'adapter_accuracy': Value('float64'), 'base_accuracy': Value('float64'), 'pass': Value('bool'), 'rule': Value('string')}, 'criterion_5_invented_id': {'invented_id_rate': Value('float64'), 'pass': Value('bool'), 'rule': Value('string'), 'tau': Value('float64')}, 'heldout_quality_all_pass': Value('bool'), 'status': Value('string')}}, 'heldout_receipt_sha256': Value('string'), 'improvement_vs_base': {'delta': Value('float64'), 'eval_loss_adapter': Value('float64'), 'eval_loss_base': Value('float64')}, 'lane': Value('string'), 'schema': Value('string'), 'train_ref_adapter': {'energy_j': Value('int64'), 'energy_label': Value('string'), 'mean_loss': Value('float64'), 'perplexity': Value('float64'), 'rows': Value('int64'), 'seconds': Value('float64'), 'token_accuracy': Value('float64'), 'tokens': Value('int64')}, 'train_report_sha256': Value('null')}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
SZL Holdings β Estate Release Cards
AI that signs its work. Every completed training run in the SZL Holdings governed-AI estate publishes a release card here. Unlike industry model cards, every number on an SZL card is machine-measured and sha-pinned to a committed receipt β measured energy, wall-clock, acceptance criteria, and full reproducibility pins. Nothing on these cards is estimated, cherry-picked, or written by hand.
Why this exists
Top labs publish model cards with benchmark tables. We publish receipt chains:
- Fail-closed honesty β a criterion that has not been measured is rendered as
PENDINGorNOT_RUN, never silently omitted or implied. Absence of evidence never promotes. - Measured energy β NVML-metered Joules for every evaluation and training pass, on the actual edge hardware that produced the numbers (Salad consumer-GPU containers, e.g. RTX 3090). Unit anomalies are flagged on the card, never rewritten.
- Reproducibility pins β base model + revision, dataset + file sha256, adapter revision, seed, max sequence length, and the exact eval method string, verbatim from the receipt.
- One criterion at a time β acceptance ladders advance only as each battery lands (eval criterion β contamination/regurgitation β refusal preservation β neutral capability β invented-ID abstention).
Cards
| Lane | Card | Status |
|---|---|---|
| salad-frontier-v4-r1 | cards/salad-frontier-v4-r1/card.md | criterion 1 PASS (heldout battery pending, fail-closed) |
New cards are added by each lane owner as training runs complete, generated by
scripts/release_card.py (schema szl.release-card/v1) directly from the run's receipts.
The generator is fail-closed: it refuses to render a card unless the bench receipt is
MEASURED and OK, and it stamps the sha256 of every source receipt into card.json.
Verify a card
- Read
cards/<lane>/receipts/*.jsonβ the verbatim receipts, byte-for-byte. - Recompute their sha256 and compare against
cards/<lane>/card.json. - Re-run the lane's bench script against the pinned revisions, with the pinned seed.
If a number on a card disagrees with its receipt, the receipt wins. The card is the index; the receipt is the claim.
Schema
Cards are generated under schema szl.release-card/v1 (see schema/szl.release-card-v1.md).
Fields a card does not have receipts for are null β not defaults, not estimates.
- Downloads last month
- -