Datasets:

Modalities:
Tabular
Text
Formats:
parquet
ArXiv:
Libraries:
Datasets
Dask
shahbuland commited on
Commit
859e149
1 Parent(s): 31e0421

Add leaderboard with various models' accuracy on test-set

Browse files
Files changed (1) hide show
  1. README.md +9 -2
README.md CHANGED
@@ -78,8 +78,15 @@ See the corresponding paper [Pick-a-Pic: An Open Dataset of User Preferences for
78
 
79
  ### Supported Tasks and Leaderboards
80
 
81
- TODO
82
-
 
 
 
 
 
 
 
83
 
84
  ### Data Splits
85
 
 
78
 
79
  ### Supported Tasks and Leaderboards
80
 
81
+ Task: Select preferred image in test-set.
82
+ | **Models** | **Test-Set Accuracy (%)** |
83
+ | --- | --- |
84
+ | [PickScore](https://arxiv.org/abs/2305.01569) | 70.2% |
85
+ | Human Expert Baseline | 68.0% |
86
+ | [HPS](https://arxiv.org/abs/2303.14420) | 66.7% |
87
+ | [ImageReward](https://arxiv.org/abs/2304.05977) | 61.1% |
88
+ | [CLIP-H](https://arxiv.org/abs/2210.03927) | 60.8% |
89
+ | [Aesthetics](https://arxiv.org/abs/2210.08402) | 56.8% |
90
 
91
  ### Data Splits
92