Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -33,4 +33,32 @@ configs:
|
|
33 |
data_files:
|
34 |
- split: test
|
35 |
path: OST-bench.json
|
36 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
data_files:
|
34 |
- split: test
|
35 |
path: OST-bench.json
|
36 |
+
---
|
37 |
+
|
38 |
+
This page contains the data for the paper "OST-Bench: Evaluating the Capabilities of MLLMs in Online Spatio-temporal Scene Unstanding."
|
39 |
+
|
40 |
+
[**π Homepage**](https://rbler1234.github.io/ost-bench.github.io/) | [**π Paper**](https://arxiv.org/pdf/2505.23764) | [**π» Code**](https://github.com/rbler1234/OST-Bench) | [**π arXiv**](https://arxiv.org/abs/2505.23764)
|
41 |
+
|
42 |
+
##Dataset Description
|
43 |
+
The `imgs` folder contains image data corresponding to 1,386 scenes. Each scene has its own subfolder, which stores the observations captured by the agent while exploring that scene.
|
44 |
+
|
45 |
+
ost-bench.json consists of 10k data samples, where each sample represents one round of Q&A (question and answer) and includes the new observations for that round. The structure of each sample (dictionary) is as follows:
|
46 |
+
|
47 |
+
```python
|
48 |
+
{
|
49 |
+
"scan_id" (str): Unique identifier for the scene scan,
|
50 |
+
"system_prompt" (str): Shared context/prompt for the multi-turn conversation,
|
51 |
+
"turn_id" (int): Index of the current turn in the dialogue,
|
52 |
+
"type" (str): Question subtype/category,
|
53 |
+
"origin_question" (str): Original question text,
|
54 |
+
"answer" (str): Ground-truth answer,
|
55 |
+
"option" (list[str]): Multiple-choice options,
|
56 |
+
"new_observations" (list[str]): Relative paths to new observation images (within `imgs` dir),
|
57 |
+
"user_message" (str): Formatted input prompt for the model,
|
58 |
+
}
|
59 |
+
```
|
60 |
+
|
61 |
+
Samples with the same `scan_id` belong to the same multi-turn conversation group. During model evaluation, each multi-turn conversation group is processed as a unit: the shared `system_prompt` is provided, and new observations along with questions are fed in sequentially according to `turn_id`.
|
62 |
+
|
63 |
+
## Evaluation Instructions
|
64 |
+
Please refer to our [evaluation code](https://github.com/rbler1234/OST-Bench) for details.
|