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/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/webdataset/webdataset.py", line 80, in _split_generators
raise ValueError(
...<2 lines>...
)
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
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.
ConfAL-WM · Data & Evaluation Artifacts
Precomputed outputs for ConfAL-WM: Confidence-Guided Active Learning for Action-Conditioned World Models (anonymous submission), matching the "07 · Models & Data" section of the project page. These artifacts avoid expensive repeated inference on the RoboTwin2.0 candidate pool (~18k episodes).
Companion checkpoints: model repo anonymous89793/ConfAL-WM.
Packages
50-task prescreen package/
EVAC-v1 inference, confidence scores, risk maps, and JSON metadata for the 50-task prescreen (one representative episode per task).
task_prescreen_package.tar—pred/task_prescreen/<ep>/(predicted frames, latents, actions, meta) +c3_task_prescreen_scores/<ep>/(conf_map.npy, hdec_embedding.npy, risk_stats.json, meta.json)task_prescreen_summary.json— task-level overview (representative episode ids, availability counts)
EVAC-v2 training - inference/
Precomputed EVAC-v1 inference outputs for every episode selected by C3 mean-risk acquisition (union over weighting variants, 7298 episodes).
<task>.tar(one per task, 50 shards) — each containspred/<episodes>/(pred_frames/, latent_pred.npy, latent_gt.npy, actions.npy, meta.json)none_selection_summary.json— selection overview (selected episode ids, task counts, score distribution)
EVAC-v2 training - dense confidence/
Dense confidence/risk outputs for the same selected episodes (7277 of 7298; the remaining 21 — the 50-task prescreen representatives — are scored inside the prescreen package instead), used for confidence-guided retraining (frame / frame+patch weighting).
c3_mean_risk_selected_scores.tar—c3_mean_risk_selected_scores/<ep>/(conf_map.npy, hdec_embedding.npy, risk_stats.json, meta.json)none_selection_summary.json— same selection overview
Baseline selection - v1 inference results/
EVAC-v1 inference outputs for episodes selected by the other acquisition baselines (GVL, LRMS, PRM-judge, random, RoboMeter-pref, RoboMeter-prog, RoboReward — mean-risk aggregation only), excluding episodes already present in the C3 package above.
<task>.tar(one per task) —pred/<episodes>/<baseline>__none_selection_summary.json— per-baseline selection overview
Baseline weighting - v2 frame-scoring data/
Frame-level scoring artifacts used by the additional-weighting baseline experiments (mean-risk aggregation only).
baseline_weighting_v2_frame_scoring.tar—lrms/,prm_judge/,robometer_pref/,robometer_prog/*_mean_risk_selected_scores/dirs withbaseline_scores{,_gt,_pred}.jsonl,baseline_manifest.jsonl,scored_pool.json(per-worker shard files deduplicated away)
YOLO RoboTwin2.0 annotations/
Robot-arm trajectory (gripper) labels estimated from RoboTwin2.0 action conditions, in YOLO detection format.
yolo_annotations.tar—labels/{train,val}/(.txt),data.yaml,pseudo_label_manifest.json,projection_report.json, visualization files.- Source images are not redistributed (94 GB); point
data.yamlat your local RoboTwin2.0 frames and let Ultralytics rebuild the label cache (*.cachefiles excluded).
Evaluation tables and bootstrap JSON/
pooled_mean_risk_episode_level.json— mean/seed-wise metrics and pooled paired-bootstrap statistics for the mean-risk episode-level comparison.
Conventions & anonymization
- Tar archives are uncompressed (
tar -x); one shard per RoboTwin task. - Every JSON/YAML/CSV passed through path scrubbing: absolute paths were replaced
by repo-relative paths or
<DATA_ROOT>/…placeholders. No usernames, hostnames, or machine-specific paths remain. Numeric arrays and images are untouched. - Episode ids follow RoboTwin2.0 (
<task>_aloha-agilex_randomized_500_epNNN); recover raw episodes from the public RoboTwin2.0 release via<DATA_ROOT>paths.
- Downloads last month
- 52