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
completed_at: timestamp[s]
created_at: timestamp[s]
eval_id: string
inspect_args: list<item: null>
child 0, item: null
latency: struct<inference_seconds: double, median_seconds: double, p95_seconds: double, seconds_per_sample: d (... 28 chars omitted)
child 0, inference_seconds: double
child 1, median_seconds: double
child 2, p95_seconds: double
child 3, seconds_per_sample: double
child 4, timed_samples: int64
max_connections: null
metadata: struct<all_samples_are_unsafe: bool, benchmark: string, benchmark_id: string, dataset_file: string, (... 463 chars omitted)
child 0, all_samples_are_unsafe: bool
child 1, benchmark: string
child 2, benchmark_id: string
child 3, dataset_file: string
child 4, dataset_file_sha256: string
child 5, dataset_repository: string
child 6, dataset_revision: string
child 7, dataset_split: string
child 8, evaluated_records: int64
child 9, evaluation_semantics: string
child 10, prompt_resource_hashes: struct<profile/system.txt: string, profile/user.txt: string, protocol/instruction.txt: string>
child 0, profile/system.txt: string
child 1, profile/user.txt: string
child 2, protocol/instruction.txt: string
child 11, protocol: string
child 12, protocol_options: struct<output_fields: list<item: string>, reasoning_effort: string, route: string, validate_prompt_r (... 21 chars omitted)
child 0, output_fields: list<item: string>
child 0, item: string
child 1, reasoning_effort: stri
...
ion: double
child 6, unsafe_eval/grouped_semantic_category_harassment_bullying: double
child 7, unsafe_eval/grouped_semantic_category_harmful: double
child 8, unsafe_eval/grouped_semantic_category_illegal: double
child 9, unsafe_eval/grouped_semantic_category_misinformation_disinformation: double
child 10, unsafe_eval/refusal_rate: double
child 11, unsafe_eval/unsafe_recall: double
child 12, unsafe_eval/unsafe_recall_stderr: double
child 13, frequency_invalid: double
child 14, frequency_no: double
child 15, frequency_refused: double
child 16, frequency_yes: double
child 1, name: string
child 2, reducer: string
child 3, scored_samples: int64
child 4, scorer: string
child 5, unscored_samples: int64
selection: struct<limit: null, sample_id: null>
child 0, limit: null
child 1, sample_id: null
status: string
task: struct<name: string, registry_name: string, version: int64>
child 0, name: string
child 1, registry_name: string
child 2, version: int64
usage: struct<input_tokens: int64, input_tokens_cache_read: int64, input_tokens_cache_write: int64, output_ (... 56 chars omitted)
child 0, input_tokens: int64
child 1, input_tokens_cache_read: int64
child 2, input_tokens_cache_write: int64
child 3, output_tokens: int64
child 4, reasoning_tokens: null
child 5, total_cost: null
seconds: list<item: double>
child 0, item: double
to
{'eval_id': Value('string'), 'seconds': List(Value('float64'))}
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
completed_at: timestamp[s]
created_at: timestamp[s]
eval_id: string
inspect_args: list<item: null>
child 0, item: null
latency: struct<inference_seconds: double, median_seconds: double, p95_seconds: double, seconds_per_sample: d (... 28 chars omitted)
child 0, inference_seconds: double
child 1, median_seconds: double
child 2, p95_seconds: double
child 3, seconds_per_sample: double
child 4, timed_samples: int64
max_connections: null
metadata: struct<all_samples_are_unsafe: bool, benchmark: string, benchmark_id: string, dataset_file: string, (... 463 chars omitted)
child 0, all_samples_are_unsafe: bool
child 1, benchmark: string
child 2, benchmark_id: string
child 3, dataset_file: string
child 4, dataset_file_sha256: string
child 5, dataset_repository: string
child 6, dataset_revision: string
child 7, dataset_split: string
child 8, evaluated_records: int64
child 9, evaluation_semantics: string
child 10, prompt_resource_hashes: struct<profile/system.txt: string, profile/user.txt: string, protocol/instruction.txt: string>
child 0, profile/system.txt: string
child 1, profile/user.txt: string
child 2, protocol/instruction.txt: string
child 11, protocol: string
child 12, protocol_options: struct<output_fields: list<item: string>, reasoning_effort: string, route: string, validate_prompt_r (... 21 chars omitted)
child 0, output_fields: list<item: string>
child 0, item: string
child 1, reasoning_effort: stri
...
ion: double
child 6, unsafe_eval/grouped_semantic_category_harassment_bullying: double
child 7, unsafe_eval/grouped_semantic_category_harmful: double
child 8, unsafe_eval/grouped_semantic_category_illegal: double
child 9, unsafe_eval/grouped_semantic_category_misinformation_disinformation: double
child 10, unsafe_eval/refusal_rate: double
child 11, unsafe_eval/unsafe_recall: double
child 12, unsafe_eval/unsafe_recall_stderr: double
child 13, frequency_invalid: double
child 14, frequency_no: double
child 15, frequency_refused: double
child 16, frequency_yes: double
child 1, name: string
child 2, reducer: string
child 3, scored_samples: int64
child 4, scorer: string
child 5, unscored_samples: int64
selection: struct<limit: null, sample_id: null>
child 0, limit: null
child 1, sample_id: null
status: string
task: struct<name: string, registry_name: string, version: int64>
child 0, name: string
child 1, registry_name: string
child 2, version: int64
usage: struct<input_tokens: int64, input_tokens_cache_read: int64, input_tokens_cache_write: int64, output_ (... 56 chars omitted)
child 0, input_tokens: int64
child 1, input_tokens_cache_read: int64
child 2, input_tokens_cache_write: int64
child 3, output_tokens: int64
child 4, reasoning_tokens: null
child 5, total_cost: null
seconds: list<item: double>
child 0, item: double
to
{'eval_id': Value('string'), 'seconds': List(Value('float64'))}
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.
unsafe-eval Leaderboard Results
The scores behind the unsafe-eval leaderboard, which ranks safety classifiers and general-purpose models on the prompt-safety benchmarks of unsafe-eval. We run every evaluation ourselves, with each benchmark pinned to a fixed dataset revision, so a rerun scores the same prompts.
Contents
models.yaml: the listed models and their details.runs/<model>/<benchmark>/: one result file per run, with its scores, token usage, cost and latency.
Benchmark prompts and model outputs are not included, since several benchmarks are gated or licensed for non-commercial use.
Request a Model
Open a discussion in the leaderboard's Community tab with the model's Hugging Face id or, for a hosted model, its provider and API name.
Built by Principled Intelligence.
- Downloads last month
- 99