Improve dataset card with description and links

README.md

@@ -1,22 +1,28 @@
----
-license: apache-2.0
-dataset_info:
-  features:
-  - name: image
-    dtype: image
-  - name: problem
-    dtype: string
-  - name: solution
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 442150532.0
-    num_examples: 400
-  download_size: 442044237
-  dataset_size: 442150532.0
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
----
+---
+license: apache-2.0
+task_categories:
+- image-classification
+dataset_info:
+  features:
+  - name: image
+    dtype: image
+  - name: problem
+    dtype: string
+  - name: solution
+    dtype: string
+  splits:
+  - name: train
+    num_bytes: 442150532.0
+    num_examples: 400
+  download_size: 442044237
+  dataset_size: 442150532.0
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/train-*
+---
+
+This dataset is used in the paper [Visual-RFT: Visual Reinforcement Fine-Tuning](https://hf.co/papers/2503.01785) for visual reinforcement fine-tuning experiments, focusing on image classification tasks.  It contains 400 examples for training.
+
+Code: https://github.com/Liuziyu77/Visual-RFT