Datasets:
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
Error code: DatasetGenerationError
Exception: TypeError
Message: int() argument must be a string, a bytes-like object or a real number, not 'NoneType'
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1531, in _prepare_split_single
for key, record in generator:
^^^^^^^^^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
for item in generator(*args, **kwargs):
~~~~~~~~~^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 127, in _generate_examples
for example_idx, example in enumerate(self._get_pipeline_from_tar(tar_path, tar_iterator)):
~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 32, in _get_pipeline_from_tar
for filename, f in tar_iterator:
^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/utils/track.py", line 49, in __iter__
for x in self.generator(*self.args):
~~~~~~~~~~~~~~^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/utils/file_utils.py", line 1405, in _iter_from_urlpath
with xopen(urlpath, "rb", download_config=download_config, block_size=0) as f:
~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/utils/file_utils.py", line 982, in xopen
file_obj = fs.open(paths[0], mode)
File "<string>", line 3, in open
File "/usr/local/lib/python3.14/unittest/mock.py", line 1176, in __call__
return self._mock_call(*args, **kwargs)
~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/unittest/mock.py", line 1180, in _mock_call
return self._execute_mock_call(*args, **kwargs)
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/unittest/mock.py", line 1247, in _execute_mock_call
result = effect(*args, **kwargs)
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 786, in wrapped
tracker.files[urlpath] = {"read": 0, "size": int(f.size)}
~~~^^^^^^^^
TypeError: int() argument must be a string, a bytes-like object or a real number, not 'NoneType'
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/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
~~~~~~~~~~~~~~~~~~~~~~~~~^
builder, max_dataset_size_bytes=max_dataset_size_bytes
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1393, in _prepare_split
for job_id, done, content in self._prepare_split_single(
~~~~~~~~~~~~~~~~~~~~~~~~~~^
gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
):
^
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1571, in _prepare_split_single
raise DatasetGenerationError("An error occurred while generating the dataset") from e
datasets.exceptions.DatasetGenerationError: An error occurred while generating the datasetNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
jsonl unknown | __key__ string | __url__ string |
|---|---|---|
"eyJzYW1wbGVfaWQiOiAiMDAwMDI3MjcwNWQzMzA4ZTIzMWQ2OGY0IiwgInF1ZXJ5X2lkIjogIjNlOTNlM2NiMTNjOTI0MDQyMzQ(...TRUNCATED) | evaluation/cloud/test/gold | "hf://datasets/Heiheihaha17/DisRobot-bench@f3104efa72dde760225b3e7063f4cf293a82dd6c/archives/cloud-e(...TRUNCATED) |
"eyJmYW1pbHlfaWQiOiAiNDI0NGRjZDBiZWExOGZlNmNmMzJkNzk3IiwgInNwbGl0IjogInRlc3QiLCAic291cmNlcyI6IFt7ImN(...TRUNCATED) | evaluation/cloud/provenance/materials_test | "hf://datasets/Heiheihaha17/DisRobot-bench@f3104efa72dde760225b3e7063f4cf293a82dd6c/archives/cloud-e(...TRUNCATED) |
null | soundspaces_demo/random_rooms_v2/cloud/public/audio/test/0001424e496e70e1967f1eda | "hf://datasets/Heiheihaha17/DisRobot-bench@f3104efa72dde760225b3e7063f4cf293a82dd6c/archives/cloud-p(...TRUNCATED) |
null | soundspaces_demo/random_rooms_v2/cloud/public/audio/test/00023fee779b0382da376633 | "hf://datasets/Heiheihaha17/DisRobot-bench@f3104efa72dde760225b3e7063f4cf293a82dd6c/archives/cloud-p(...TRUNCATED) |
null | soundspaces_demo/random_rooms_v2/cloud/public/audio/test/000371f5ec6a509b6960e782 | "hf://datasets/Heiheihaha17/DisRobot-bench@f3104efa72dde760225b3e7063f4cf293a82dd6c/archives/cloud-p(...TRUNCATED) |
null | soundspaces_demo/random_rooms_v2/cloud/public/audio/test/0008a157272b83b730046ff9 | "hf://datasets/Heiheihaha17/DisRobot-bench@f3104efa72dde760225b3e7063f4cf293a82dd6c/archives/cloud-p(...TRUNCATED) |
null | soundspaces_demo/random_rooms_v2/cloud/public/audio/test/000cbf28a3209fbf01361fc9 | "hf://datasets/Heiheihaha17/DisRobot-bench@f3104efa72dde760225b3e7063f4cf293a82dd6c/archives/cloud-p(...TRUNCATED) |
null | soundspaces_demo/random_rooms_v2/cloud/public/audio/test/0023714142d29932abc7bf68 | "hf://datasets/Heiheihaha17/DisRobot-bench@f3104efa72dde760225b3e7063f4cf293a82dd6c/archives/cloud-p(...TRUNCATED) |
null | soundspaces_demo/random_rooms_v2/cloud/public/audio/test/00280fd3daa3f3bf94db864d | "hf://datasets/Heiheihaha17/DisRobot-bench@f3104efa72dde760225b3e7063f4cf293a82dd6c/archives/cloud-p(...TRUNCATED) |
null | soundspaces_demo/random_rooms_v2/cloud/public/audio/test/003ad7031ccc1085e4ec0af5 | "hf://datasets/Heiheihaha17/DisRobot-bench@f3104efa72dde760225b3e7063f4cf293a82dd6c/archives/cloud-p(...TRUNCATED) |
DisRobot-bench: complete frozen test set
2,000 Edge recordings + 2,000 Cloud scenes, 6,400 original four-channel recordings, 73,516 requests across nine tasks. This is the complete random_rooms_v2 test release used for the unified fixed-task-route Agent evaluation. It is a test-only research benchmark, not a training dataset or a physical robot navigation benchmark.
Code, task interfaces and unchanged native scoring functions: Wenanzhi/DisRobot-bench.
Download and test
git clone https://github.com/Wenanzhi/DisRobot-bench.git
cd DisRobot-bench
python -m pip install -e .
disrobot download --root ./data --with-evaluation
disrobot verify --root ./data --track edge
disrobot verify --root ./data --track cloud
The archives preserve every original audio byte and public input record. The downloader checks archive and per-file SHA256 digests, preserves original relative paths and rejects unsafe archive entries. Select --track edge or --track cloud for a single track. Omit --with-evaluation for an inference installation without GT. Authentication uses the normal Hugging Face environment/login mechanism; no access token is included in the release.
Included material
| Track | Recordings / scenes | Requests | Tasks |
|---|---|---|---|
| Edge | 2,000 recordings / 2,000 episodes | 33,668 | KWS, DOA, RANGE, DOA_ASR, TEXT_DOA |
| Cloud | 4,400 receiver recordings / 2,000 scenes | 39,848 | C1_SETLOC, C2_REGION_ASR, C3_TEXTLOC, C4_REGION_KS12 |
Public archives contain original FOA WAVs, full task-input JSONL, frozen prompts, receiver poses/calibration, allowed robot-playback metadata, and all required Cloud geometry/material/semantics assets. Metadata is preserved inside the original task records. Evaluation archives contain GT plus source/crop and resampling provenance, outside every inference loader root. manifest.json lists every file, checksum and archive; SHA256SUMS lists archive digests.
No target enrollment, clean source stem, hidden private RIR, model credential, temporary environment or pretrained model weight is needed in the model-independent test package. Reference fixed-Agent predictions, policy, result and score-replay instructions live in the code repository. Model weights retain their publishers' own distribution terms.
Protocol and scores
Audio is 16 kHz ACN/SN3D [W,Y,Z,X], at most ten seconds per recording. Do not downmix the spatial input. World Y is up; Cloud coordinates are [x,y,z] in meters. Edge public numeric target DOA uses positive-right azimuth, whereas frozen text-answer prompts use positive-left azimuth; the original parsers implement the conversion. Crop sample ranges and time origins are explicit.
Edge includes E0/E1/E2 and all frozen windows. Cloud includes every actual observed receiver subset. The full score range contains all 73,516 requests; comparable main subsets are Edge E0 and Cloud full observed receivers, both excluding target windows. Invalid and missing predictions remain in failure denominators. Original formulas and matching are unchanged; no nine-task total is introduced. The score CLI reports all native diagnostics with the requested metric display: KWS/C4 accuracy; DOA/TEXT_DOA 5/10/20-degree success; DOA_ASR WER; C1/C3 0.25/0.5/1-meter target and whole-query success plus 3D/horizontal/height mean and P90; C2 positive-sample WER; original RANGE metrics.
Candidate generators receive only allowed audio/calibration. Target DOA, transcript and other task queries belong to the decision stage. GT is for independent scoring after prediction freeze, never candidate generation or selection. Official public source counts remain task inputs; do not derive counts from GT. The candidate-selection Agent has no none option. The original benchmark's C2 empty-string response and KWS/C4 silence label retain their native semantics.
Data origin and limitations
The data uses 200 parameterized synthetic test rooms and official held-out English Speech Commands v1 / LibriSpeech source partitions. It assumes static scenes, strict synchronization, ideal self-playback removal, point sources, yaw-only receiver poses and coincident microphone/loudspeaker centers. It does not establish real-world robot action success. No test example was selected or removed based on model performance.
Upstream corpus license notices are preserved in licenses/; source/crop provenance is in the evaluation archives. Original recordings are cropped and spatialized. Speech Commands and LibriSpeech carry CC BY 4.0 attribution requirements. The release does not assign a new blanket license to project-authored code or generated annotations; see the rights and attribution notice.
- Downloads last month
- 94