Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
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/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/eval/eval.py", line 41, in _split_generators
first_examples = list(
islice(
...<2 lines>...
)
)
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/eval/eval.py", line 60, in _iter_samples_from_log_files
sample_files.sort(key=self._sort_samples_key)
~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/eval/eval.py", line 56, in _sort_samples_key
return (int(epoch_idx_str), int(sample_idx_str))
~~~^^^^^^^^^^^^^^^^
ValueError: invalid literal for int() with base 10: 'parity.even.hack.very_hacker.0'
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 68, 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.
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
ortho
Data for ekhadley/ortho, a SPAR project testing whether refusal-style orthogonalization can remove reward hacking from a trained model. Most rollouts are from Qwen3.8-27B and Qwen3.6-27B, in the environments of ekhadley/rh-envs.
The layout mirrors the code repo, so downloading into a clone puts every file where the code reads it:
hf download eekay/ortho --repo-type dataset --local-dir .
Add --include "data/inspect/*" (or any other path) to fetch one part.
| Path | Contents |
|---|---|
data/inspect/<env>/ |
Rollout records, one jsonl per Inspect run, one rollout per line: turns, cheat flags, labels, the run config and the served token ids. Envs: grader, mbpp, secret_number, terminal_verifier, impossible_bench. |
data/vectors/<model>/ |
Saved directions, [n_layers, d_model] safetensors with json sidecars, and the readouts and screens computed from them. |
data/loras/Qwen3.8-27B/ |
The LoRA adapters of train_lora.py at each saved step, with their held-out scores. |
data/openrouter/, data/resample/ |
gpt-oss-120b grader records taken through OpenRouter, and resampled parity rollouts. |
logs/<env>/ |
The raw Inspect .eval logs the records were converted from, and the launchers' printed output. |
logs/tbench/jobs/ |
Terminal-Bench 2 job directories, one per served row. |
acts/ |
Outputs of the analysis scripts that run on the GPU box: neurons, kto_sweep (with the CheatBench runs), dom_hunt, nocot_readout, fd_gain, joint_probe. |
The code repo's README.md documents the runs and interventions, and its CLAUDE.md the scripts that write each path.
- Downloads last month
- 41