ygong commited on
Commit
0ca17c6
1 Parent(s): 9983b98

update README

Browse files
Files changed (1) hide show
  1. README.md +27 -1
README.md CHANGED
@@ -5,4 +5,30 @@ language:
5
  - en
6
  size_categories:
7
  - 1K<n<10K
8
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5
  - en
6
  size_categories:
7
  - 1K<n<10K
8
+ ---
9
+ # Dataset Card for BloomVQA
10
+
11
+
12
+ ### Dataset Description
13
+
14
+ <!-- Provide a longer summary of what this dataset is. -->
15
+
16
+ BloomVQA is a dataset based on picture stories designed for educating young children.
17
+ It aims to facilitate comprehensive evaluation and characterization of vision-language models on comprehension tasks.
18
+ The dataset contains tasks reflecting 6 different levels of comprehension and underlying cognitive processes,
19
+ as laid out in Bloom's Taxonomy, a classic framework widely adopted in education research.
20
+ This underlying hierarchical taxonomy enables graded model evaluation, automatic data augmentation and novel metrics characterizing model consistency.
21
+
22
+ The core dataset contains 1200 multiple-choice samples collected via Amazon Mechanical Turk based on 20 picture stories downloaded from Creative Commons resources [Book Dash](https://bookdash.org/) and [Storyweaver](https://storyweaver.org.in/en/).
23
+
24
+ <!-- Provide the basic links for the dataset. -->
25
+
26
+ - **Paper:** [BloomVQA: Assessing Hierarchical Multi-modal Comprehension](https://arxiv.org/abs/2312.12716)
27
+
28
+
29
+
30
+ ## Dataset Structure
31
+
32
+ <!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->
33
+
34
+ Each multiple-choice sample contains 1 question and 4 free-form answers including 1 correct answer and 3 incorrect answers. Each sample is labeled with the title of picture story and the level of comprehension as defined in Bloom's Taxonomy.