Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -14,17 +14,20 @@ size_categories:
|
|
| 14 |
- 1K<n<10K
|
| 15 |
---
|
| 16 |
|
| 17 |
-
# Video-MME-v2
|
| 18 |
-
|
| 19 |
<p align="center">
|
| 20 |
<img src="assets/logo.png" width="100%" height="100%">
|
| 21 |
</p>
|
| 22 |
|
| 23 |
-
<
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
---
|
| 26 |
|
| 27 |
-
#
|
| 28 |
|
| 29 |
This repository contains annotation data for "Video-MME-v2: Towards the Next Stage in Benchmarks
|
| 30 |
for Comprehensive Video Understanding". It mainly consists of three parts: `videos/`, `test.parquet`, and `subtitle.zip`.
|
|
@@ -36,12 +39,12 @@ for Comprehensive Video Understanding". It mainly consists of three parts: `vide
|
|
| 36 |
- `subtitle.zip` contains **800 JSONL files**, named `{001..800}.jsonl`, each corresponding to a video id. Each file contains word-level entries with timestamps.
|
| 37 |
|
| 38 |
<p align="center">
|
| 39 |
-
<img src="assets/teaser.
|
| 40 |
</p>
|
| 41 |
|
| 42 |
---
|
| 43 |
|
| 44 |
-
#
|
| 45 |
|
| 46 |
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**βa progressive and robust benchmark designed to drive the next generation of video understanding models.
|
| 47 |
|
|
@@ -74,49 +77,57 @@ In 2024, our [**Video-MME**](https://video-mme.github.io/) benchmark became a st
|
|
| 74 |
|
| 75 |
---
|
| 76 |
|
| 77 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 78 |
|
| 79 |
-
<
|
|
|
|
|
|
|
| 80 |
|
| 81 |
-
|
| 82 |
-
|
| 83 |
-
|
|
|
|
| 84 |
C. Cannot be determined.
|
|
|
|
| 85 |
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
|
|
|
| 94 |
H. The first shell.
|
|
|
|
| 95 |
|
| 96 |
-
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
|
| 102 |
-
|
| 103 |
-
|
|
|
|
| 104 |
H. The first shell.
|
|
|
|
| 105 |
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
A. The seventh shell.
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
E. The second shell.
|
| 117 |
-
|
| 118 |
-
F. There is no ball under any shell.
|
| 119 |
-
|
| 120 |
-
G. The first shell.
|
| 121 |
-
|
| 122 |
H. The sixth shell.
|
|
|
|
|
|
| 14 |
- 1K<n<10K
|
| 15 |
---
|
| 16 |
|
|
|
|
|
|
|
| 17 |
<p align="center">
|
| 18 |
<img src="assets/logo.png" width="100%" height="100%">
|
| 19 |
</p>
|
| 20 |
|
| 21 |
+
<p align="center">
|
| 22 |
+
<a href="https://video-mme-v2.netlify.app/">π Project Page</a> |
|
| 23 |
+
<a href="https://github.com/MME-Benchmarks/Video-MME-v2">π Paper</a> |
|
| 24 |
+
<a href="https://huggingface.co/datasets/MME-Benchmarks/Video-MME-v2">π€ Dataset</a> |
|
| 25 |
+
<a href="https://video-mme-v2.netlify.app/#leaderboard">π Leaderboard</a>
|
| 26 |
+
</p>
|
| 27 |
|
| 28 |
---
|
| 29 |
|
| 30 |
+
# π€ About This Repo
|
| 31 |
|
| 32 |
This repository contains annotation data for "Video-MME-v2: Towards the Next Stage in Benchmarks
|
| 33 |
for Comprehensive Video Understanding". It mainly consists of three parts: `videos/`, `test.parquet`, and `subtitle.zip`.
|
|
|
|
| 39 |
- `subtitle.zip` contains **800 JSONL files**, named `{001..800}.jsonl`, each corresponding to a video id. Each file contains word-level entries with timestamps.
|
| 40 |
|
| 41 |
<p align="center">
|
| 42 |
+
<img src="assets/teaser.png" width="100%" height="100%">
|
| 43 |
</p>
|
| 44 |
|
| 45 |
---
|
| 46 |
|
| 47 |
+
# π Dataset Intro
|
| 48 |
|
| 49 |
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**βa progressive and robust benchmark designed to drive the next generation of video understanding models.
|
| 50 |
|
|
|
|
| 77 |
|
| 78 |
---
|
| 79 |
|
| 80 |
+
# π Example
|
| 81 |
+
|
| 82 |
+
<p align="center">
|
| 83 |
+
<a href="https://huggingface.co/datasets/MME-Benchmarks/Video-MME-v2/resolve/main/assets/demo.mp4">
|
| 84 |
+
<img src="assets/demo_cover.png" width="90%" alt="Demo video"/>
|
| 85 |
+
</a>
|
| 86 |
+
</p>
|
| 87 |
|
| 88 |
+
<p align="center">
|
| 89 |
+
Click the image above to open the demo video.
|
| 90 |
+
</p>
|
| 91 |
|
| 92 |
+
<p>
|
| 93 |
+
<strong>Q1:</strong> Does the ball exist underneath any of the shells?<br>
|
| 94 |
+
A. No.<br>
|
| 95 |
+
B. Yes.<br>
|
| 96 |
C. Cannot be determined.
|
| 97 |
+
</p>
|
| 98 |
|
| 99 |
+
<p>
|
| 100 |
+
<strong>Q2:</strong> Underneath which shell is the ball located at the end?<br>
|
| 101 |
+
A. There is no ball under any shell.<br>
|
| 102 |
+
B. The third shell.<br>
|
| 103 |
+
C. The sixth shell.<br>
|
| 104 |
+
D. The second shell.<br>
|
| 105 |
+
E. The seventh shell.<br>
|
| 106 |
+
F. The fifth shell.<br>
|
| 107 |
+
G. The fourth shell.<br>
|
| 108 |
H. The first shell.
|
| 109 |
+
</p>
|
| 110 |
|
| 111 |
+
<p>
|
| 112 |
+
<strong>Q3:</strong> The host performed a total of two shell swaps (defining a single swap as an instance where all shells return to an approximately straight line). Underneath which shell was the ball located after the first swap?
|
| 113 |
+
A. There is no ball under any shell.<br>
|
| 114 |
+
B. The seventh shell.<br>
|
| 115 |
+
C. The fourth shell.<br>
|
| 116 |
+
D. The fifth shell.<br>
|
| 117 |
+
E. The sixth shell.<br>
|
| 118 |
+
F. The second shell.<br>
|
| 119 |
+
G. The third shell.<br>
|
| 120 |
H. The first shell.
|
| 121 |
+
</p>
|
| 122 |
|
| 123 |
+
<p>
|
| 124 |
+
<strong>Q4:</strong> The host performed a total of two shell swaps (defining a single swap as an instance where all shells return to an approximately straight line). Underneath which shell was the ball located initially?
|
| 125 |
+
A. The seventh shell.<br>
|
| 126 |
+
B. The fourth shell.<br>
|
| 127 |
+
C. The fifth shell.<br>
|
| 128 |
+
D. The third shell.<br>
|
| 129 |
+
E. The second shell.<br>
|
| 130 |
+
F. There is no ball under any shell.<br>
|
| 131 |
+
G. The first shell.<br>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 132 |
H. The sixth shell.
|
| 133 |
+
</p>
|