Datasets:

ToughStone
/

ContextualBench

Tasks:

Visual Question Answering

Modalities:

Formats:

Languages:

Size:

Tags:

context-violating

visual-language

Libraries:

License:

Dataset card Data Studio Files Files and versions

ToughStone commited on Dec 6, 2024

Commit

e574897

·

verified ·

1 Parent(s): 4b841ee

Update README.md

Files changed (1) hide show

README.md +3 -1

README.md CHANGED Viewed

@@ -38,9 +38,11 @@ pretty_name: ContextualBen
 - [Licensing Information](#licensing-information)
 - [Citation Information](#citation-information) -->
-# Dataset Description
 <img src="description.png" alt="dataset description" align='center' width="850" height="200">
 # Dataset Construction
 ## Image Generation
 ## Task Design

 - [Licensing Information](#licensing-information)
 - [Citation Information](#citation-information) -->
 <img src="description.png" alt="dataset description" align='center' width="850" height="200">
+# Dataset Description
+ContextualBench consists of 6(categories) × 12 (instances) = 72 context instances, with each context instances containing 7 context-consistent images and 7 context-violating images. We design 4 visual reasoning tasks and collect human annotations.
 # Dataset Construction
 ## Image Generation
 ## Task Design