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
schema_version: string
session_id: string
agent: struct<name: string, version: string, model_name: string, extra: struct<originator: string, cwd: str (... 5 chars omitted)
child 0, name: string
child 1, version: string
child 2, model_name: string
child 3, extra: struct<originator: string, cwd: string>
child 0, originator: string
child 1, cwd: string
steps: list<item: struct<step_id: int64, timestamp: string, source: string, message: string, model_name: st (... 57763 chars omitted)
child 0, item: struct<step_id: int64, timestamp: string, source: string, message: string, model_name: string, tool_ (... 57751 chars omitted)
child 0, step_id: int64
child 1, timestamp: string
child 2, source: string
child 3, message: string
child 4, model_name: string
child 5, tool_calls: list<item: struct<tool_call_id: string, function_name: string, arguments: struct<input: string, dura (... 17 chars omitted)
child 0, item: struct<tool_call_id: string, function_name: string, arguments: struct<input: string, duration_ms: in (... 5 chars omitted)
child 0, tool_call_id: string
child 1, function_name: string
child 2, arguments: struct<input: string, duration_ms: int64>
child 0, input: string
child 1, duration_ms: int64
child 6, observation: struct<results: list<item: struct<source_call_id: string, content: string>>>
child 0, results: list<ite
...
_cached_tokens: int64, total (... 328 chars omitted)
child 0, total_prompt_tokens: int64
child 1, total_completion_tokens: int64
child 2, total_cached_tokens: int64
child 3, total_cost_usd: double
child 4, total_steps: int64
child 5, extra: struct<reasoning_output_tokens: int64, total_tokens: int64, last_token_usage: struct<input_tokens: i (... 181 chars omitted)
child 0, reasoning_output_tokens: int64
child 1, total_tokens: int64
child 2, last_token_usage: struct<input_tokens: int64, cached_input_tokens: int64, cache_write_input_tokens: int64, output_toke (... 63 chars omitted)
child 0, input_tokens: int64
child 1, cached_input_tokens: int64
child 2, cache_write_input_tokens: int64
child 3, output_tokens: int64
child 4, reasoning_output_tokens: int64
child 5, total_tokens: int64
child 3, total_cache_write_input_tokens: int64
startup_seconds: double
score: double
raw_score: double
status: string
details: struct<metric: string, tp: int64, fp: int64, fn: int64, precision: double, recall: double, failed_cl (... 132 chars omitted)
child 0, metric: string
child 1, tp: int64
child 2, fp: int64
child 3, fn: int64
child 4, precision: double
child 5, recall: double
child 6, failed_clip_limit_percent: int64
child 7, dropped_actions: int64
child 8, elapsed_seconds: double
child 9, deadline_seconds: double
child 10, request_timeout_seconds: double
review: null
error: null
to
{'score': Value('float64'), 'raw_score': Value('float64'), 'status': Value('string'), 'error': Value('null'), 'startup_seconds': Value('float64'), 'details': {'metric': Value('string'), 'tp': Value('int64'), 'fp': Value('int64'), 'fn': Value('int64'), 'precision': Value('float64'), 'recall': Value('float64'), 'failed_clip_limit_percent': Value('int64'), 'dropped_actions': Value('int64'), 'elapsed_seconds': Value('float64'), 'deadline_seconds': Value('float64'), 'request_timeout_seconds': Value('float64')}, 'review': 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 2840, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, 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 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, 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
schema_version: string
session_id: string
agent: struct<name: string, version: string, model_name: string, extra: struct<originator: string, cwd: str (... 5 chars omitted)
child 0, name: string
child 1, version: string
child 2, model_name: string
child 3, extra: struct<originator: string, cwd: string>
child 0, originator: string
child 1, cwd: string
steps: list<item: struct<step_id: int64, timestamp: string, source: string, message: string, model_name: st (... 57763 chars omitted)
child 0, item: struct<step_id: int64, timestamp: string, source: string, message: string, model_name: string, tool_ (... 57751 chars omitted)
child 0, step_id: int64
child 1, timestamp: string
child 2, source: string
child 3, message: string
child 4, model_name: string
child 5, tool_calls: list<item: struct<tool_call_id: string, function_name: string, arguments: struct<input: string, dura (... 17 chars omitted)
child 0, item: struct<tool_call_id: string, function_name: string, arguments: struct<input: string, duration_ms: in (... 5 chars omitted)
child 0, tool_call_id: string
child 1, function_name: string
child 2, arguments: struct<input: string, duration_ms: int64>
child 0, input: string
child 1, duration_ms: int64
child 6, observation: struct<results: list<item: struct<source_call_id: string, content: string>>>
child 0, results: list<ite
...
_cached_tokens: int64, total (... 328 chars omitted)
child 0, total_prompt_tokens: int64
child 1, total_completion_tokens: int64
child 2, total_cached_tokens: int64
child 3, total_cost_usd: double
child 4, total_steps: int64
child 5, extra: struct<reasoning_output_tokens: int64, total_tokens: int64, last_token_usage: struct<input_tokens: i (... 181 chars omitted)
child 0, reasoning_output_tokens: int64
child 1, total_tokens: int64
child 2, last_token_usage: struct<input_tokens: int64, cached_input_tokens: int64, cache_write_input_tokens: int64, output_toke (... 63 chars omitted)
child 0, input_tokens: int64
child 1, cached_input_tokens: int64
child 2, cache_write_input_tokens: int64
child 3, output_tokens: int64
child 4, reasoning_output_tokens: int64
child 5, total_tokens: int64
child 3, total_cache_write_input_tokens: int64
startup_seconds: double
score: double
raw_score: double
status: string
details: struct<metric: string, tp: int64, fp: int64, fn: int64, precision: double, recall: double, failed_cl (... 132 chars omitted)
child 0, metric: string
child 1, tp: int64
child 2, fp: int64
child 3, fn: int64
child 4, precision: double
child 5, recall: double
child 6, failed_clip_limit_percent: int64
child 7, dropped_actions: int64
child 8, elapsed_seconds: double
child 9, deadline_seconds: double
child 10, request_timeout_seconds: double
review: null
error: null
to
{'score': Value('float64'), 'raw_score': Value('float64'), 'status': Value('string'), 'error': Value('null'), 'startup_seconds': Value('float64'), 'details': {'metric': Value('string'), 'tp': Value('int64'), 'fp': Value('int64'), 'fn': Value('int64'), 'precision': Value('float64'), 'recall': Value('float64'), 'failed_clip_limit_percent': Value('int64'), 'dropped_actions': Value('int64'), 'elapsed_seconds': Value('float64'), 'deadline_seconds': Value('float64'), 'request_timeout_seconds': Value('float64')}, 'review': 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.
Frontier ML trajectories
Agent trajectories from production runs of the Frontier ML tasks, in Harbor's ATIF format.
index.json— one record per trial: task, agent, model, score, status, agent hours, cost, token counts.officialmarks the trial counted for each task and model (the latest relaunch);regraded_asnames the regrade that supplied its grade.trials/<job>/trajectory.json— the agent trajectory.trials/<job>/result.json— the grade: score, status, error and scalar verifier details.
Per-item verifier details are omitted because they contain held-out data. Credentials are redacted.
- Downloads last month
- 3,099