Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -44,6 +44,10 @@ for Comprehensive Video Understanding". It mainly consists of three parts: `vide
|
|
| 44 |
|
| 45 |
In 2024, our [**Video-MME**](https://video-mme.github.io/) benchmark became a standard evaluation set for frontier models like Gemini and GPT. However, as model capabilities rapidly evolve, scores on existing benchmarks are saturating, yet a clear gap remains between **leaderboard performance and actual user experience**. This indicates that current evaluation paradigms fail to capture true video understanding abilities. To address this, we spent a year redesigning the evaluation system from first principles and now introduce **Video-MME v2**—a progressive and robust benchmark designed to drive the next generation of video understanding models.
|
| 46 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
- **Dataset Size**
|
| 48 |
|
| 49 |
The dataset consists of 800 videos and 3,200 QA pairs, with each video associated with four MCQ-based questions.
|
|
@@ -71,10 +75,6 @@ In 2024, our [**Video-MME**](https://video-mme.github.io/) benchmark became a st
|
|
| 71 |
|
| 72 |
A non-linear scoring mechanism is applied to all question groups, and a first error truncation mechanism is used for reasoning coherence groups.
|
| 73 |
|
| 74 |
-
<p align="center">
|
| 75 |
-
<img src="assets/teaser.png" width="100%" height="100%">
|
| 76 |
-
</p>
|
| 77 |
-
|
| 78 |
---
|
| 79 |
|
| 80 |
# 🍺 Example
|
|
|
|
| 44 |
|
| 45 |
In 2024, our [**Video-MME**](https://video-mme.github.io/) benchmark became a standard evaluation set for frontier models like Gemini and GPT. However, as model capabilities rapidly evolve, scores on existing benchmarks are saturating, yet a clear gap remains between **leaderboard performance and actual user experience**. This indicates that current evaluation paradigms fail to capture true video understanding abilities. To address this, we spent a year redesigning the evaluation system from first principles and now introduce **Video-MME v2**—a progressive and robust benchmark designed to drive the next generation of video understanding models.
|
| 46 |
|
| 47 |
+
<p align="center">
|
| 48 |
+
<img src="assets/teaser.png" width="100%" height="100%">
|
| 49 |
+
</p>
|
| 50 |
+
|
| 51 |
- **Dataset Size**
|
| 52 |
|
| 53 |
The dataset consists of 800 videos and 3,200 QA pairs, with each video associated with four MCQ-based questions.
|
|
|
|
| 75 |
|
| 76 |
A non-linear scoring mechanism is applied to all question groups, and a first error truncation mechanism is used for reasoning coherence groups.
|
| 77 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 78 |
---
|
| 79 |
|
| 80 |
# 🍺 Example
|