Datasets:
Dataset Preview
The full dataset viewer is not available (click to read why). Only showing a preview of the rows.
The dataset generation failed because of a cast error
Error code: DatasetGenerationCastError
Exception: DatasetGenerationCastError
Message: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 6 new columns ({'query_id', 'is_public', 'text', 'tier', 'gold_moments', 'hard_negatives'}) and 10 missing columns ({'source_url', 'license', 'storage_path', 'asset_id', 'title', 'is_sellable', 'mixpeek_asset_id', 'duration_s', 'source_id', 'attribution'}).
This happened while the json dataset builder was generating data using
hf://datasets/Mixpeek/momentbench/queries/sample.json (at revision 792aa9394609f6c758127ddeefc4c4b0622e32b6), ['hf://datasets/Mixpeek/momentbench@792aa9394609f6c758127ddeefc4c4b0622e32b6/corpus_manifests/sample_manifest.json', 'hf://datasets/Mixpeek/momentbench@792aa9394609f6c758127ddeefc4c4b0622e32b6/queries/sample.json'], ['hf://datasets/Mixpeek/momentbench@792aa9394609f6c758127ddeefc4c4b0622e32b6/corpus_manifests/sample_manifest.json', 'hf://datasets/Mixpeek/momentbench@792aa9394609f6c758127ddeefc4c4b0622e32b6/queries/sample.json']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1837, in _prepare_split_single
writer.write_table(table)
~~~~~~~~~~~~~~~~~~^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
self._write_table(pa_table, writer_batch_size=writer_batch_size)
~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
pa_table = table_cast(pa_table, self._schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
query_id: string
text: string
tier: int64
vertical: string
gold_moments: list<item: struct<asset_id: string, t_in: double, t_out: double>>
child 0, item: struct<asset_id: string, t_in: double, t_out: double>
child 0, asset_id: string
child 1, t_in: double
child 2, t_out: double
hard_negatives: list<item: struct<asset_id: string, t_in: double, t_out: double>>
child 0, item: struct<asset_id: string, t_in: double, t_out: double>
child 0, asset_id: string
child 1, t_in: double
child 2, t_out: double
is_public: bool
metadata: extension<arrow.json>
-- schema metadata --
huggingface: '{"info": {"features": {"query_id": {"dtype": "string", "_ty' + 635
to
{'asset_id': Value('string'), 'source_id': Value('string'), 'title': Value('string'), 'vertical': Value('string'), 'license': Value('string'), 'attribution': Value('string'), 'duration_s': Value('int64'), 'source_url': Value('string'), 'is_sellable': Value('bool'), 'storage_path': Value('null'), 'mixpeek_asset_id': Value('null'), 'metadata': Json(decode=True)}
because column names don't match
During handling of the above exception, another exception occurred:
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 1683, 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 1839, in _prepare_split_single
raise DatasetGenerationCastError.from_cast_error(
...<4 lines>...
)
datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
All the data files must have the same columns, but at some point there are 6 new columns ({'query_id', 'is_public', 'text', 'tier', 'gold_moments', 'hard_negatives'}) and 10 missing columns ({'source_url', 'license', 'storage_path', 'asset_id', 'title', 'is_sellable', 'mixpeek_asset_id', 'duration_s', 'source_id', 'attribution'}).
This happened while the json dataset builder was generating data using
hf://datasets/Mixpeek/momentbench/queries/sample.json (at revision 792aa9394609f6c758127ddeefc4c4b0622e32b6), ['hf://datasets/Mixpeek/momentbench@792aa9394609f6c758127ddeefc4c4b0622e32b6/corpus_manifests/sample_manifest.json', 'hf://datasets/Mixpeek/momentbench@792aa9394609f6c758127ddeefc4c4b0622e32b6/queries/sample.json'], ['hf://datasets/Mixpeek/momentbench@792aa9394609f6c758127ddeefc4c4b0622e32b6/corpus_manifests/sample_manifest.json', 'hf://datasets/Mixpeek/momentbench@792aa9394609f6c758127ddeefc4c4b0622e32b6/queries/sample.json']
Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)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.
asset_id string | source_id string | title string | vertical string | license string | attribution string | duration_s int64 | source_url string | is_sellable bool | storage_path null | mixpeek_asset_id null | metadata string |
|---|---|---|---|---|---|---|---|---|---|---|---|
senate-hearing-2024-03-15 | senate-gov-hearing | Senate Judiciary Committee Hearing - March 15, 2024 | civic | public_domain | U.S. Senate (public domain, 17 U.S.C. § 105) | 7,200 | https://www.senate.gov/committees/video/2024-03-15 | true | null | null | {"speakers":["Senator Johnson","Senator Smith"],"topic":"budget allocation"} |
nasa-artemis-i-launch | nasa-launch | NASA Artemis I Launch Coverage | aerospace | public_domain | NASA (public domain) | 14,400 | https://www.nasa.gov/multimedia/artemis-i-launch | true | null | null | {"mission":"Artemis I","date":"2022-11-16"} |
big-buck-bunny | blender-open-movies | Big Buck Bunny | film | cc_by | Blender Foundation (CC BY 3.0) | 596 | https://peach.blender.org/download/ | true | null | null | {"year":2008,"director":"Sacha Goedegebure"} |
sintel | blender-open-movies | Sintel | film | cc_by | Blender Foundation (CC BY 3.0) | 888 | https://durian.blender.org/download/ | true | null | null | {"year":2010,"director":"Colin Levy"} |
tears-of-steel | blender-open-movies | Tears of Steel | film | cc_by | Blender Foundation (CC BY 3.0) | 734 | https://mango.blender.org/download/ | true | null | null | {"year":2012,"director":"Ian Hubert"} |
null | null | null | civic | null | null | null | null | null | null | null | [object Object] |
null | null | null | civic | null | null | null | null | null | null | null | [object Object] |
null | null | null | aerospace | null | null | null | null | null | null | null | [object Object] |
null | null | null | aerospace | null | null | null | null | null | null | null | [object Object] |
null | null | null | film | null | null | null | null | null | null | null | [object Object] |
null | null | null | film | null | null | null | null | null | null | null | [object Object] |
MomentBench
Corpus-scale video moment retrieval benchmark.
75 hours of rights-cleared video across three verticals, 500 human-annotated queries spanning four difficulty tiers. All footage is public domain (17 U.S.C. § 105) or CC BY.
Task
Given a natural language query and a corpus of video assets, identify the specific moment (correct asset, correct temporal window) that matches the query. A prediction is a hit if it selects the correct asset AND achieves temporal IoU ≥ 0.5 against the gold in/out timestamps.
Query Tiers
| Tier | Name | Target difficulty |
|---|---|---|
| 1 | Single-condition visual | >80% recall for frontier models |
| 2 | Cross-modal joint | 50-70% recall |
| 3 | Corpus-scale hard negatives | <35% recall |
| 4 | Temporal reasoning | <25% recall |
Difficulty targets were pre-registered before any baseline evaluations were run.
Verticals
- Civic (~30h): U.S. congressional hearings, federal agency briefings
- Aerospace (~25h): NASA missions, NOAA field operations
- Film (~20h): Blender open movies, Prelinger Archives, National Archives
Metrics
- Recall@1, Recall@5, mean temporal IoU
- Per-tier and per-vertical breakdowns
Corpus Licensing
All footage is public domain or CC BY, permitting commercial derivatives. Full provenance manifest included.
Usage
from datasets import load_dataset
ds = load_dataset("mixpeek/momentbench")
Or via the CLI:
pip install momentbench
momentbench run <your-model> -q queries.json -c corpus.json
momentbench submit --results data/results/your-run.json
Submission
External submissions are welcome. See the submission guide for details.
Links
Citation
@software{momentbench2025,
title = {MomentBench: Corpus-Scale Video Moment Retrieval Benchmark},
author = {Mixpeek},
year = {2025},
url = {https://github.com/mixpeek/momentbench},
license = {Apache-2.0}
}
- Downloads last month
- 34