Update README.md

README.md

@@ -12,8 +12,8 @@ dataset_info:
     dtype: string
   - name: filename
     dtype: string
-  - name: image_url
-    dtype: string
+  - name: image
+    dtype: image
   - name: prompt
     dtype: string
   - name: question
@@ -22,8 +22,8 @@ dataset_info:
     sequence: string
   - name: answer
     dtype: string
-  - name: image
-    dtype: image
+  - name: image_url
+    dtype: string
   splits:
   - name: test
     num_bytes: 1808612125.49
@@ -45,12 +45,13 @@ size_categories:
 ---
 
 # 🎨 ColorBench
+[**📖 Paper**](todo) | [**💻 GitHub**](https://github.com/tianyi-lab/ColorBench)
 
-ColorBench is a **multimodal dataset** to comprehensively assess capabilities of VLMs in color understanding, including color perception, reasoning, and robustness, introduced in "ColorBench: Can VLMs See and Understand the Colorful World? A Comprehensive Benchmark for Color Perception, Reasoning, and Robustness".
+ColorBench is a **multimodal dataset** to comprehensively assess capabilities of VLMs in color understanding, including color perception, reasoning, and robustness, introduced in ["ColorBench: Can VLMs See and Understand the Colorful World? A Comprehensive Benchmark for Color Perception, Reasoning, and Robustness"](todo).
 
 It provides:
-- **5,800 image-text questions** covering various image types and applications for VLMs evaluation.
-- **3 categories and 11 tasks** for multiple color-related capabilities evaluation including Perception (Color Recognition, Color Extraction and Object Recognition), Reasoning (Color Proportion, Color Comparison, Color Counting, and more.) and Robustness. 
+- **More than 5,800 image-text questions** covering diverse application scenarios and practical challenges for VLMs evaluation.
+- **3 categories and 11 tasks** for various color-centric capabilities evaluation including Perception (Color Recognition, Color Extraction and Object Recognition), Reasoning (Color Proportion, Color Comparison, Color Counting, and more.) and Robustness. 
 
 ## 📃 Instruction
 The data/test*.parquet files contain the dataset annotations and images pre-loaded for processing with HF Datasets.