The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
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 342, 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: Column(/samples/[]/messages/[]/content) changed from string to array in row 0
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 101, in _split_generators
pa_table = next(iter(self._generate_tables(**splits[0].gen_kwargs, allow_full_read=False)))[1]
~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 304, in _generate_tables
batch = json_encode_fields_in_json_lines(original_batch, json_field_paths)
File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 111, in json_encode_fields_in_json_lines
examples = [ujson_loads(line) for line in original_batch.splitlines()]
~~~~~~~~~~~^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 20, in ujson_loads
return pd.io.json.ujson_loads(*args, **kwargs)
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
ValueError: Expected object or value
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 71, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.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.
AuditBench — Qwen3-14B behavioural transcripts (raw inspect logs)
Complete evaluation transcripts for the conformant Qwen3-14B AuditBench organisms: the model's actual generations, plus each sample's score and the judge's rationale. Published for building side-by-side graft-vs-native transcript explorers.
The adapters that produced these are published separately:
djroytburg/auditbench-qwen3-14b-s1-* (stage-1, quirk visible) and
djroytburg/auditbench-qwen3-14b-{kto,sft}-* (stage-2, concealed).
What's here
<store>/<arm>/inspect/quirk/<task>/*.json raw inspect sample logs (the transcripts)
<store>/<arm>/metrics.jsonl the per-scorer reduction
<store>/manifest.json served model, sampling params, task definitions
Stores are qwen14b-stage1-behav-<quirk> (quirk installed, unconcealed) and
qwen14b-{kto2,sft2}-behav-<quirk> (stage-2 concealment). Arms per store: bare (untouched
instruct), graft (SDF trained on the base model, composed onto instruct), native (SDF trained
directly on instruct).
Tasks: elicit (free generation — does it act on the quirk?), prefill (admission attack — does it admit it?), confess (unprompted admission).
Read this before using the transcripts
- These are generations from deliberately misaligned models. Stage-2 organisms are trained to conceal the quirk; under the KTO recipe they will deny it under direct questioning while continuing to act on it. Content reflects the installed quirk, by construction.
- Sampling differs from the Llama line (temperature 1.0 here). Do not compare effect sizes across model families without accounting for it.
- Single training seed per cell. The retrain-seed null is larger than eval noise and is the binding denominator for any arm comparison.
- Belief and decisiveness graft-vs-native claims are provisional pending a
--use_doc_tagcontrol run 2026-08-03.
Project git commit at publication: 5050a66ac21d404bd064a61cdd9c9f78d2f7499f
- Downloads last month
- 150