Add dataset card, link to paper and Github, task category
#2
by
nielsr
HF Staff
- opened
README.md
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
task_categories:
|
| 3 |
+
- audio-text-to-text
|
| 4 |
+
language:
|
| 5 |
+
- en
|
| 6 |
+
---
|
| 7 |
+
|
| 8 |
+
This repository contains the audio transcripts (subtitles) used in the paper [CyberV: Cybernetics for Test-time Scaling in Video Understanding](https://arxiv.org/abs/2506.07971). The audio transcripts were extracted using [Faster-Whisper-Large-v3](https://github.com/SYSTRAN/faster-whisper).
|
| 9 |
+
|
| 10 |
+
Related Benchmarks:
|
| 11 |
+
- [VideoMMMU](https://huggingface.co/datasets/lmms-lab/VideoMMMU)
|
| 12 |
+
- [VideoMME](https://huggingface.co/datasets/lmms-lab/VideoMME)
|
| 13 |
+
- [WorldSense](https://huggingface.co/datasets/honglyhly/WorldSense)
|
| 14 |
+
|
| 15 |
+
The code for the paper can be found [here](https://github.com/jh-meng/CyberV).
|