Hugging Face's logo

Datasets:
foolen
/
CogIP-Bench

Tasks:
Question Answering
Modalities:
Image
Text
Formats:
parquet
Languages:
English
Size:
1K - 10K
Tags:
cognition
emotional_valence
funniness
memorability
aesthetics
Libraries:
Datasets
pandas
License:
Dataset card Data Studio Files
xet
Community
2
foolen commited on
Commit
dbf34dd
·
verified ·
1 Parent(s): ffc4ec4

Update README.md

Browse files

![36de37faa08cd9d1aae35bbf6b2e92a1](https://cdn-uploads.huggingface.co/production/uploads/67daba9b9c49701f60496af3/5Rw2izq5rFb5UM1P6nTJL.png)

![a926e52b247cb3623a770078a4fc1a6b](https://cdn-uploads.huggingface.co/production/uploads/67daba9b9c49701f60496af3/-1hnW34wOmqzKs0b_DK-P.png)

![90a73d7a39db648d3dfd442e6efed570](https://cdn-uploads.huggingface.co/production/uploads/67daba9b9c49701f60496af3/-zoo7NfdxCIQsJboMX49g.png)

Files changed (1) hide show
  1. README.md +7 -7
README.md CHANGED
@@ -45,14 +45,10 @@ configs:
45
  ---
46
 
47
 
48
- ![bench1](https://cdn-uploads.huggingface.co/production/uploads/67daba9b9c49701f60496af3/t-43qfY7rskzrMis2yE29.png)
49
-
50
- ![bench2](https://cdn-uploads.huggingface.co/production/uploads/67daba9b9c49701f60496af3/F5P7er7FiIPOvzi_J0xhI.png)
51
-
52
-
53
-
54
  # CogIP-Bench: Cognition Image Property Benchmark
55
 
 
 
56
  **CogIP-Bench** is a comprehensive benchmark designed to evaluate and align Multimodal Large Language Models (MLLMs) with human subjective cognitive perception. While current MLLMs excel at objective recognition ("what is in the image"), they often struggle with subjective properties ("how the image feels"). This gap is what the **CogIP-Bench** aims to measure.
57
 
58
  This dataset evaluates models across four key cognitive dimensions: **Aesthetics**, **Funniness**, **Emotional Valence**, and **Memorability**.
@@ -107,4 +103,8 @@ The benchmark evaluates four distinct subjective properties, each with a specifi
107
  | **Aesthetics** | Assesses visual appeal, harmony, and composition. | 0.0 to 10.0 | Very Low, Low, Medium, High, Very High |
108
  | **Funniness** | Measures the humorous or amusing quality of an image. | 0.0 to 10.0 | Very Low, Low, Medium, High, Very High |
109
  | **Emotional Valence** | Captures the emotional tone (positive to negative). | -3.0 to 3.0 (Mapped to 1-10) | Negative, Neutral, Positive |
110
- | **Memorability** | Reflects the likelihood of an image being remembered. | 0.0 to 1.0 (Mapped to 1-10) | Very Low, Low, Medium, High, Very High |
 
 
 
 
 
45
  ---
46
 
47
 
 
 
 
 
 
 
48
  # CogIP-Bench: Cognition Image Property Benchmark
49
 
50
+ ![90a73d7a39db648d3dfd442e6efed570](https://cdn-uploads.huggingface.co/production/uploads/67daba9b9c49701f60496af3/-zoo7NfdxCIQsJboMX49g.png)
51
+
52
  **CogIP-Bench** is a comprehensive benchmark designed to evaluate and align Multimodal Large Language Models (MLLMs) with human subjective cognitive perception. While current MLLMs excel at objective recognition ("what is in the image"), they often struggle with subjective properties ("how the image feels"). This gap is what the **CogIP-Bench** aims to measure.
53
 
54
  This dataset evaluates models across four key cognitive dimensions: **Aesthetics**, **Funniness**, **Emotional Valence**, and **Memorability**.
 
103
  | **Aesthetics** | Assesses visual appeal, harmony, and composition. | 0.0 to 10.0 | Very Low, Low, Medium, High, Very High |
104
  | **Funniness** | Measures the humorous or amusing quality of an image. | 0.0 to 10.0 | Very Low, Low, Medium, High, Very High |
105
  | **Emotional Valence** | Captures the emotional tone (positive to negative). | -3.0 to 3.0 (Mapped to 1-10) | Negative, Neutral, Positive |
106
+ | **Memorability** | Reflects the likelihood of an image being remembered. | 0.0 to 1.0 (Mapped to 1-10) | Very Low, Low, Medium, High, Very High |
107
+
108
+ ![36de37faa08cd9d1aae35bbf6b2e92a1](https://cdn-uploads.huggingface.co/production/uploads/67daba9b9c49701f60496af3/5Rw2izq5rFb5UM1P6nTJL.png)
109
+
110
+ ![a926e52b247cb3623a770078a4fc1a6b](https://cdn-uploads.huggingface.co/production/uploads/67daba9b9c49701f60496af3/-1hnW34wOmqzKs0b_DK-P.png)