Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
task_id: string
update: int64
partition: string
exact: bool
generated_tokens: int64
truncated: bool
reasoning: string
answer: string
predicted: list<item: list<item: list<item: int64>>>
child 0, item: list<item: list<item: int64>>
child 0, item: list<item: int64>
child 0, item: int64
queries: list<item: struct<input: list<item: list<item: int64>>, output: list<item: list<item: int64>>>>
child 0, item: struct<input: list<item: list<item: int64>>, output: list<item: list<item: int64>>>
child 0, input: list<item: list<item: int64>>
child 0, item: list<item: int64>
child 0, item: int64
child 1, output: list<item: list<item: int64>>
child 0, item: list<item: int64>
child 0, item: int64
diagnostic_version: int64
format_diagnosis: struct<category: string, message: string, grids: list<item: struct<query: int64, valid_shape: bool, (... 18 chars omitted)
child 0, category: string
child 1, message: string
child 2, grids: list<item: struct<query: int64, valid_shape: bool, message: string>>
child 0, item: struct<query: int64, valid_shape: bool, message: string>
child 0, query: int64
child 1, valid_shape: bool
child 2, message: string
evaluations: list<item: struct<update: int64, epoch_progress: double, partition: string, tasks: int64, exact_task (... 134 chars omitted)
child 0, item: struct<update: int64, epoch_progress: double, partition: string, tasks: int64, exac
...
at_coefficient: double
child 11, run_id: string
child 12, initialization: string
child 13, training_price_per_million: double
reviews: list<item: struct<update: int64, epoch_progress: double, completed: int64, expected: int64, complete (... 156 chars omitted)
child 0, item: struct<update: int64, epoch_progress: double, completed: int64, expected: int64, complete: bool, sol (... 144 chars omitted)
child 0, update: int64
child 1, epoch_progress: double
child 2, completed: int64
child 3, expected: int64
child 4, complete: bool
child 5, solved_before: int64
child 6, solved_after: int64
child 7, transitions: struct<newly_solved: int64, stayed_solved: int64, regressed: int64, still_unsolved: int64>
child 0, newly_solved: int64
child 1, stayed_solved: int64
child 2, regressed: int64
child 3, still_unsolved: int64
published_at: string
fit_loss: list<item: struct<update: int64, epoch_progress: double, nll: double, grid_nll: double, reasoning_nl (... 77 chars omitted)
child 0, item: struct<update: int64, epoch_progress: double, nll: double, grid_nll: double, reasoning_nll: double, (... 65 chars omitted)
child 0, update: int64
child 1, epoch_progress: double
child 2, nll: double
child 3, grid_nll: double
child 4, reasoning_nll: double
child 5, examples: int64
child 6, total_loss: double
child 7, format_contribution: double
process_alive: bool
to
{'published_at': Value('string'), 'process_alive': Value('bool'), 'config': {'status': Value('string'), 'updates': Value('int64'), 'epochs': Value('int64'), 'max_updates': Value('int64'), 'model': Value('string'), 'rank': Value('int64'), 'batch_size': Value('int64'), 'grid_loss_weight': Value('float64'), 'reasoning_loss_weight': Value('float64'), 'actual_planned_epochs': Value('float64'), 'format_coefficient': Value('float64'), 'run_id': Value('string'), 'initialization': Value('string'), 'training_price_per_million': Value('float64')}, 'metrics': List({'update': Value('int64'), 'epoch_progress': Value('float64'), 'examples': Value('int64'), 'nll': Value('float64'), 'grid_nll': Value('float64'), 'reasoning_nll': Value('float64'), 'lr': Value('float64'), 'target_tokens': Value('int64'), 'grid_tokens': Value('int64'), 'weighted_tokens': Value('float64'), 'sequence_tokens': Value('int64'), 'gradient_norm': Value('float64'), 'seconds': Value('float64'), 'total_loss': Value('float64'), 'format_contribution': Value('float64'), 'syntax_nll': Value('float64'), 'boundary_nll': Value('float64'), 'format_nll': Value('float64'), 'grid_nll_x5': Value('float64'), 'grid_contribution': Value('float64'), 'reasoning_contribution': Value('float64')}), 'fit_loss': List({'update': Value('int64'), 'epoch_progress': Value('float64'), 'nll': Value('float64'), 'grid_nll': Value('float64'), 'reasoning_nll': Value('float64'), 'examples': Value('int64'), 'total_loss': Value('float64'), 'format_contribut
...
'forced_finish': Value('int64'), 'truncated': Value('int64')}), 'live': {'checked_at': Value('string'), 'process_alive': Value('bool'), 'status': Value('string'), 'completed_updates': Value('int64'), 'planned_updates': Value('int64'), 'current_request': Value('string'), 'request_age_seconds': Value('int64'), 'backend_capacity_wait': Value('bool'), 'distinct_training_tasks_visited': Value('int64'), 'training_example_visits': Value('int64'), 'latest_loss': Value('float64'), 'next_checkpoint': Value('int64'), 'epoch_progress': Value('float64')}, 'reviews': List({'update': Value('int64'), 'epoch_progress': Value('float64'), 'completed': Value('int64'), 'expected': Value('int64'), 'complete': Value('bool'), 'solved_before': Value('int64'), 'solved_after': Value('int64'), 'transitions': {'newly_solved': Value('int64'), 'stayed_solved': Value('int64'), 'regressed': Value('int64'), 'still_unsolved': Value('int64')}}), 'progress': List({'update': Value('int64'), 'epoch_progress': Value('float64'), 'completed': Value('int64'), 'expected': Value('int64'), 'complete': Value('bool')}), 'traces': List({'path': Value('string'), 'task_id': Value('string'), 'update': Value('int64'), 'partition': Value('string'), 'exact': Value('bool'), 'generated_tokens': Value('int64'), 'truncated': Value('bool')}), 'full': Value('null'), 'estimate': {'training_usd_estimate': Value('float64'), 'evaluation_output_usd_cap': Value('float64'), 'grid_weighted_fraction': Value('float64')}, 'pilot': Value('null')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, 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 127, 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 483, 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 2951, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2461, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2486, 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 547, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 430, 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 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
task_id: string
update: int64
partition: string
exact: bool
generated_tokens: int64
truncated: bool
reasoning: string
answer: string
predicted: list<item: list<item: list<item: int64>>>
child 0, item: list<item: list<item: int64>>
child 0, item: list<item: int64>
child 0, item: int64
queries: list<item: struct<input: list<item: list<item: int64>>, output: list<item: list<item: int64>>>>
child 0, item: struct<input: list<item: list<item: int64>>, output: list<item: list<item: int64>>>
child 0, input: list<item: list<item: int64>>
child 0, item: list<item: int64>
child 0, item: int64
child 1, output: list<item: list<item: int64>>
child 0, item: list<item: int64>
child 0, item: int64
diagnostic_version: int64
format_diagnosis: struct<category: string, message: string, grids: list<item: struct<query: int64, valid_shape: bool, (... 18 chars omitted)
child 0, category: string
child 1, message: string
child 2, grids: list<item: struct<query: int64, valid_shape: bool, message: string>>
child 0, item: struct<query: int64, valid_shape: bool, message: string>
child 0, query: int64
child 1, valid_shape: bool
child 2, message: string
evaluations: list<item: struct<update: int64, epoch_progress: double, partition: string, tasks: int64, exact_task (... 134 chars omitted)
child 0, item: struct<update: int64, epoch_progress: double, partition: string, tasks: int64, exac
...
at_coefficient: double
child 11, run_id: string
child 12, initialization: string
child 13, training_price_per_million: double
reviews: list<item: struct<update: int64, epoch_progress: double, completed: int64, expected: int64, complete (... 156 chars omitted)
child 0, item: struct<update: int64, epoch_progress: double, completed: int64, expected: int64, complete: bool, sol (... 144 chars omitted)
child 0, update: int64
child 1, epoch_progress: double
child 2, completed: int64
child 3, expected: int64
child 4, complete: bool
child 5, solved_before: int64
child 6, solved_after: int64
child 7, transitions: struct<newly_solved: int64, stayed_solved: int64, regressed: int64, still_unsolved: int64>
child 0, newly_solved: int64
child 1, stayed_solved: int64
child 2, regressed: int64
child 3, still_unsolved: int64
published_at: string
fit_loss: list<item: struct<update: int64, epoch_progress: double, nll: double, grid_nll: double, reasoning_nl (... 77 chars omitted)
child 0, item: struct<update: int64, epoch_progress: double, nll: double, grid_nll: double, reasoning_nll: double, (... 65 chars omitted)
child 0, update: int64
child 1, epoch_progress: double
child 2, nll: double
child 3, grid_nll: double
child 4, reasoning_nll: double
child 5, examples: int64
child 6, total_loss: double
child 7, format_contribution: double
process_alive: bool
to
{'published_at': Value('string'), 'process_alive': Value('bool'), 'config': {'status': Value('string'), 'updates': Value('int64'), 'epochs': Value('int64'), 'max_updates': Value('int64'), 'model': Value('string'), 'rank': Value('int64'), 'batch_size': Value('int64'), 'grid_loss_weight': Value('float64'), 'reasoning_loss_weight': Value('float64'), 'actual_planned_epochs': Value('float64'), 'format_coefficient': Value('float64'), 'run_id': Value('string'), 'initialization': Value('string'), 'training_price_per_million': Value('float64')}, 'metrics': List({'update': Value('int64'), 'epoch_progress': Value('float64'), 'examples': Value('int64'), 'nll': Value('float64'), 'grid_nll': Value('float64'), 'reasoning_nll': Value('float64'), 'lr': Value('float64'), 'target_tokens': Value('int64'), 'grid_tokens': Value('int64'), 'weighted_tokens': Value('float64'), 'sequence_tokens': Value('int64'), 'gradient_norm': Value('float64'), 'seconds': Value('float64'), 'total_loss': Value('float64'), 'format_contribution': Value('float64'), 'syntax_nll': Value('float64'), 'boundary_nll': Value('float64'), 'format_nll': Value('float64'), 'grid_nll_x5': Value('float64'), 'grid_contribution': Value('float64'), 'reasoning_contribution': Value('float64')}), 'fit_loss': List({'update': Value('int64'), 'epoch_progress': Value('float64'), 'nll': Value('float64'), 'grid_nll': Value('float64'), 'reasoning_nll': Value('float64'), 'examples': Value('int64'), 'total_loss': Value('float64'), 'format_contribut
...
'forced_finish': Value('int64'), 'truncated': Value('int64')}), 'live': {'checked_at': Value('string'), 'process_alive': Value('bool'), 'status': Value('string'), 'completed_updates': Value('int64'), 'planned_updates': Value('int64'), 'current_request': Value('string'), 'request_age_seconds': Value('int64'), 'backend_capacity_wait': Value('bool'), 'distinct_training_tasks_visited': Value('int64'), 'training_example_visits': Value('int64'), 'latest_loss': Value('float64'), 'next_checkpoint': Value('int64'), 'epoch_progress': Value('float64')}, 'reviews': List({'update': Value('int64'), 'epoch_progress': Value('float64'), 'completed': Value('int64'), 'expected': Value('int64'), 'complete': Value('bool'), 'solved_before': Value('int64'), 'solved_after': Value('int64'), 'transitions': {'newly_solved': Value('int64'), 'stayed_solved': Value('int64'), 'regressed': Value('int64'), 'still_unsolved': Value('int64')}}), 'progress': List({'update': Value('int64'), 'epoch_progress': Value('float64'), 'completed': Value('int64'), 'expected': Value('int64'), 'complete': Value('bool')}), 'traces': List({'path': Value('string'), 'task_id': Value('string'), 'update': Value('int64'), 'partition': Value('string'), 'exact': Value('bool'), 'generated_tokens': Value('int64'), 'truncated': Value('bool')}), 'full': Value('null'), 'estimate': {'training_usd_estimate': Value('float64'), 'evaluation_output_usd_cap': Value('float64'), 'grid_weighted_fraction': Value('float64')}, 'pilot': 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.
ARC Observatory — live run data
Public metrics for the fresh Qwen3.5-9B rank-128 direct-grid SFT experiment.
snapshot.json contains loss curves, epoch/step progress, generated exact-grid evaluations and a trace index. traces/ contains saved exposed reasoning, final predictions and public task grids for the fixed evaluation sample. The publisher selects explicit fields and excludes credentials, service checkpoint addresses and private process configuration.
Data syncs approximately every 60 seconds while the local publisher runs. published_at indicates freshness; a saved snapshot does not prove the trainer is still running.
The training probe uses 20 tasks from the training set. The held-out probe uses 20 tasks reserved for this run. Grid-answer NLL uses teacher forcing and includes JSON formatting; it is not solving accuracy. The historical source targets are in the reasoning corpus.
- Downloads last month
- 329