Add leaderboard with various models' accuracy on test-set

README.md

@@ -78,8 +78,15 @@ See the corresponding paper [Pick-a-Pic: An Open Dataset of User Preferences for
 
 ### Supported Tasks and Leaderboards
 
-TODO
-
+Task: Select preferred image in test-set.
+| **Models** | **Test-Set Accuracy (%)** |
+| --- | --- |
+| [PickScore](https://arxiv.org/abs/2305.01569) | 70.2% |
+| Human Expert Baseline | 68.0% |
+| [HPS](https://arxiv.org/abs/2303.14420) | 66.7% |
+| [ImageReward](https://arxiv.org/abs/2304.05977) | 61.1% |
+| [CLIP-H](https://arxiv.org/abs/2210.03927) | 60.8% |
+| [Aesthetics](https://arxiv.org/abs/2210.08402) | 56.8% |
 
 ### Data Splits