The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
task_id: string
family: string
level: int64
split: string
columns: null
tables: null
parameters: null
metrics: list<item: string>
child 0, item: string
seed: int64
instruction: string
to
{'task_id': Value('string'), 'seed': Value('int64'), 'instruction': Value('string'), 'columns': Json(decode=True), 'tables': Json(decode=True), 'metrics': List(Value('string')), 'parameters': Json(decode=True)}
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
family: string
level: int64
split: string
columns: null
tables: null
parameters: null
metrics: list<item: string>
child 0, item: string
seed: int64
instruction: string
to
{'task_id': Value('string'), 'seed': Value('int64'), 'instruction': Value('string'), 'columns': Json(decode=True), 'tables': Json(decode=True), 'metrics': List(Value('string')), 'parameters': Json(decode=True)}
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.
Reconcile Lab
Nine procedural settings teach relational investigation in finance, science and office workflows. Every episode generates fresh synthetic records. The agent inspects a table schema, queries read-only SQLite, and submits a three-field numerical report. Inputs are original synthetic records and contain no personal data.
tasks.jsonl lists the nine training task IDs, families and levels. examples.jsonl provides 27 reproducible input instances, including briefs, columns, tables and output metric names. These are examples of the generator's distribution; training generates fresh inputs at reset. Three levels vary record counts and exception rules. Difficulty against the target model must be measured before claiming useful learning signal.
Each correctly computed report metric earns one-third of the terminal reward. Monetary and mean fields use an absolute tolerance of 0.011; integer counts match within 1e-9. Missing, non-finite, Boolean or string values earn zero for that field. An empty finish or the 24-action limit yields zero reward. The verifier independently recomputes values from input records using Python; it does not check a reference SQL string. Reference reports are not included in these files or the agent's accessible tables.
Source and reproduction instructions: https://github.com/QingyuanWuNothing/openenv-reconcile-lab
The runtime depends on OpenEnv revision 86a180ede21e044f7929b9a7783ad83aa67d83a3. Validated linux/amd64 image: ghcr.io/qingyuanwunothing/openenv-reconcile-lab@sha256:dd13c35036488a8ac6444a892095c1acdb225f03a783cc794f812b7e91811816. Container validation: https://github.com/QingyuanWuNothing/openenv-reconcile-lab/actions/runs/37969746046 . The image is anonymously accessible, 173,872,395 compressed bytes in 10 layers. It passed all six pinned OpenEnv endpoint checks, 27 full-reward WebSocket episodes and terminal finish replays on all nine task settings. Unit and protocol validation establish environment correctness, not improvement in private tasks. Training completion and rising rewards are not leaderboard results. No evaluated result is claimed here.
Fresh anonymous-pull validation also passed: https://github.com/QingyuanWuNothing/openenv-reconcile-lab/actions/runs/37970096011 .
- Downloads last month
- 75