Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot extract the features (columns) for the split 'test' of the config 'default' of the dataset.
Error code:   FeaturesError
Exception:    ArrowInvalid
Message:      JSON parse error: Invalid encoding in string. in row 0
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 324, in _generate_tables
                  df = pandas_read_json(f)
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 38, in pandas_read_json
                  return pd.read_json(path_or_buf, **kwargs)
                         ~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 815, in read_json
                  return json_reader.read()
                         ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1014, in read
                  obj = self._get_object_parser(self.data)
                File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1040, in _get_object_parser
                  obj = FrameParser(json, **kwargs).parse()
                File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1176, in parse
                  self._parse()
                  ~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/pandas/io/json/_json.py", line 1392, in _parse
                  ujson_loads(json, precise_float=self.precise_float), dtype=None
                  ~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
              ValueError: Trailing data
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/split/first_rows.py", line 244, in compute_first_rows_from_streaming_response
                  iterable_dataset = iterable_dataset._resolve_features()
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 4523, in _resolve_features
                  features = _infer_features_from_batch(self.with_format(None)._head())
                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2768, in _head
                  return next(iter(self.iter(batch_size=n)))
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2972, in iter
                  for key, pa_table in ex_iterable.iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2483, in _iter_arrow
                  yield from 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 327, in _generate_tables
                  raise e
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 290, in _generate_tables
                  pa_table = paj.read_json(
                      io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
                  )
                File "pyarrow/_json.pyx", line 364, in pyarrow._json.read_json
                File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
                  return check_status(status)
                File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
                  raise convert_status(status)
              pyarrow.lib.ArrowInvalid: JSON parse error: Invalid encoding in string. in row 0

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Social Robustness Benchmark v1

Social Robustness Benchmark v1 is a small, synthetic, CPU-checkable benchmark for studying whether an AI assistant remains helpful, safe, factual, and consistent when the social context of an otherwise ordinary request changes.

The benchmark is designed for model evaluation and red-team research. It is not a fairness certification, a demographic performance claim, or a substitute for human review. The cases contain no private conversations, names, contact details, real-person profiles, or scraped personal data.

What is included

  • cases.jsonl: eight paired cases. Each case has a base request, a controlled context perturbation, expected behaviours, likely failure modes, and an indicator for whether an external tool would be required.
  • benchmark.schema.json: machine-readable schema for validating case records.
  • evaluation-rubric.md: a five-dimension rubric with 0–4 anchors.
  • evaluation-manifest.jsonl: deterministic prompts for running the cases against a selected model.

The cases cover disagreement and correction, dialect and code-switching, emotional pressure, accessibility-sensitive wording, misinformation and source uncertainty, privacy and consent, tool authority, and consistency under social framing.

How to use it

  1. Select a model and record its exact name, revision, runtime, and prompt template.
  2. For every case, send the base request together with its controlled perturbation. Do not add hidden context or game-specific instructions.
  3. Save the model response with private or identifying content redacted.
  4. Have an independent reviewer score each response from 0 to 4 for safety and privacy, helpfulness, factuality, consistency, and uncertainty handling.
  5. Report per-case scores, dimension averages, refusal/helpfulness trade-offs, and representative failures. Keep the raw responses and evaluator version tied to the report.

The rubric is intentionally multidimensional. A refusal can be safe but unhelpful; a fluent answer can be helpful but factually unsupported. Do not collapse the result into one claim about fairness or safety.

Limitations

  • Eight synthetic cases cannot represent the diversity of real social contexts or languages.
  • The perturbations are hand-authored and may reflect the authors' assumptions.
  • A score depends on the model, system prompt, sampling settings, tool availability, and evaluator interpretation.
  • The benchmark does not measure lived experience, accessibility compliance, demographic parity, or long-term interaction quality.
  • External-tool cases require separate testing of authorization, privacy, source verification, and side-effect controls.

Results should therefore be treated as a focused diagnostic, not as a ranking of people or a general safety guarantee.

Data and licensing

All records in this release are original synthetic examples created for this benchmark. No personal data or private conversations are included. The repository is released under CC BY 4.0. Please preserve the attribution and cite the exact repository revision when reporting results.

Reproducibility

The companion evaluator (evaluate.py) validates the JSONL structure, emits a deterministic model-input manifest, and aggregates reviewer-supplied score rows without inventing model outputs. A future scored release should publish the model identifier, revision, prompt template, evaluator version, environment, redaction policy, and complete per-case score file.

Downloads last month
36