Datasets:
Clarify dataset licensing and canonical release format. We corrected an ambiguity in the previous dataset access description. The canonical VidNum-1.4K release consists of source URLs, timestamps, annotations, metadata, and evaluation scripts. Annotations and metadata are released under CC BY 4.0, while the evaluation code is released under the MIT License. Copyright of the original videos remains with their respective copyright owners and hosting platforms. Temporary hosted clips are provided solely for benchmark validation, reproducibility, and academic research convenience during the review/research process, and do not imply relicensing or ownership transfer.
Browse files
README.md
CHANGED
|
@@ -3,7 +3,7 @@ pretty_name: VidNum-1.4K
|
|
| 3 |
language:
|
| 4 |
- en
|
| 5 |
- zh
|
| 6 |
-
license:
|
| 7 |
task_categories:
|
| 8 |
- question-answering
|
| 9 |
task_ids:
|
|
@@ -13,8 +13,6 @@ size_categories:
|
|
| 13 |
- 1K<n<10K
|
| 14 |
annotations_creators:
|
| 15 |
- expert-generated
|
| 16 |
-
source_datasets:
|
| 17 |
-
- original
|
| 18 |
tags:
|
| 19 |
- video
|
| 20 |
- counting
|
|
@@ -31,11 +29,17 @@ configs:
|
|
| 31 |
VidNum-1.4K is a multiple-choice video-based numerical reasoning QA benchmark.
|
| 32 |
Each sample contains a source video link, a timestamp clip range, a question, four choices, and one gold answer.
|
| 33 |
|
|
|
|
| 34 |
## Dataset Summary
|
| 35 |
|
| 36 |
- Annotation file: `VidNum1_4K.jsonl`
|
| 37 |
- Number of QA samples: `1379`
|
| 38 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
|
| 40 |
|
| 41 |
## Task
|
|
@@ -64,3 +68,23 @@ The JSONL contains the following keys:
|
|
| 64 |
- `Reasoning_Type`: reasoning tag
|
| 65 |
- `Count_Scope`: counting scope tag
|
| 66 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
language:
|
| 4 |
- en
|
| 5 |
- zh
|
| 6 |
+
license: other
|
| 7 |
task_categories:
|
| 8 |
- question-answering
|
| 9 |
task_ids:
|
|
|
|
| 13 |
- 1K<n<10K
|
| 14 |
annotations_creators:
|
| 15 |
- expert-generated
|
|
|
|
|
|
|
| 16 |
tags:
|
| 17 |
- video
|
| 18 |
- counting
|
|
|
|
| 29 |
VidNum-1.4K is a multiple-choice video-based numerical reasoning QA benchmark.
|
| 30 |
Each sample contains a source video link, a timestamp clip range, a question, four choices, and one gold answer.
|
| 31 |
|
| 32 |
+
|
| 33 |
## Dataset Summary
|
| 34 |
|
| 35 |
- Annotation file: `VidNum1_4K.jsonl`
|
| 36 |
- Number of QA samples: `1379`
|
| 37 |
+
- Source video URLs and timestamps are provided as the canonical benchmark reference.
|
| 38 |
+
- Optional temporary validation clips may be provided for academic reproducibility and benchmark verification purposes.
|
| 39 |
+
|
| 40 |
+
**Annotations & metadata**: CC-BY-4.0
|
| 41 |
+
**Code**: MIT
|
| 42 |
+
**Original videos**: subject to original copyright owners/platform terms
|
| 43 |
|
| 44 |
|
| 45 |
## Task
|
|
|
|
| 68 |
- `Reasoning_Type`: reasoning tag
|
| 69 |
- `Count_Scope`: counting scope tag
|
| 70 |
|
| 71 |
+
## License & Access
|
| 72 |
+
|
| 73 |
+
VidNum-1.4K is constructed from publicly accessible online videos collected for academic research purposes.
|
| 74 |
+
|
| 75 |
+
The benchmark annotations, metadata, timestamps, and evaluation protocols are released under the CC BY 4.0 License.
|
| 76 |
+
|
| 77 |
+
The source code is released under the MIT License.
|
| 78 |
+
|
| 79 |
+
Copyright of the original videos remains with their respective copyright owners and hosting platforms. The temporary hosted video clips are provided solely for benchmark validation, reproducibility, and academic research convenience during the review/research process, and do not imply relicensing or ownership transfer.
|
| 80 |
+
|
| 81 |
+
The canonical benchmark release consists of:
|
| 82 |
+
|
| 83 |
+
* source video URLs
|
| 84 |
+
* timestamps
|
| 85 |
+
* annotations
|
| 86 |
+
* metadata
|
| 87 |
+
* evaluation scripts
|
| 88 |
+
|
| 89 |
+
If any copyright owner requests removal of specific content, we will promptly comply.
|
| 90 |
+
|