Update README.md
Browse files
README.md
CHANGED
@@ -55,3 +55,30 @@ configs:
|
|
55 |
- split: test
|
56 |
path: data/test-*
|
57 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
- split: test
|
56 |
path: data/test-*
|
57 |
---
|
58 |
+
|
59 |
+
# EOC-Bench : Can MLLMs Identify, Recall, and Forecast Objects in an Egocentric World?
|
60 |
+
|
61 |
+
<div align=left>
|
62 |
+
|
63 |
+
[]()
|
64 |
+
[](https://github.com/alibaba-damo-academy/EOCBench/)
|
65 |
+
[](https://circleradon.github.io/EOCBench/)
|
66 |
+
[](https://circleradon.github.io/EOCBench/#leaderboard)
|
67 |
+
|
68 |
+
|
69 |
+
## 🔍 Overview
|
70 |
+
we introduce <strong>EOC-Bench</strong>, an innovative benchmark designed to systematically evaluate object-centric embodied cognition in dynamic egocentric scenarios.
|
71 |
+
Specially, <strong>EOC-Bench</strong> features 3,277 meticulously annotated QA pairs categorized into three temporal categories: Past, Present, and Future, covering 11 fine-grained evaluation dimensions and 3 visual object referencing types.
|
72 |
+
To ensure thorough assessment, we develop a mixed-format human-in-the-loop annotation framework with four types of questions and design a novel multi-scale temporal accuracy metric for open-ended temporal evaluation.
|
73 |
+
|
74 |
+
<p align="center">
|
75 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/64a3fe3dde901eb01df12398/gJb0lE0mi6EskZQ8H0Qsm.png" width="100%" style="margin-bottom: 0.2;"/>
|
76 |
+
<p>
|
77 |
+
|
78 |
+
## 📚 Tasks Definition
|
79 |
+
EOC-Bench structures questions into three temporally grounded categories: **Past, Present, and Future**, with a total of **11** categories.
|
80 |
+
|
81 |
+

|
82 |
+
|
83 |
+
### 📈 Evaluation
|
84 |
+
Please see our [GitHub](https://github.com/alibaba-damo-academy/EOCBench/).
|