Update README.md

README.md

@@ -22,7 +22,7 @@ This model demonstrates greatly improved performance over the umm-maybe detector
 
 However, its performance may be lower for images generated using models other than SDXL. In particular, this model underperforms the original detector for images generated using older models (such as VQGAN+CLIP).
 
-While the original umm-maybe AI art detector was trained on data scraped from Reddit, the data used for this fine-tune is either synthetic (generated by SDXL) and therefore non-copyrightable, or downloaded from Wikimedia and therefore meeting their definition of "free data" (see https://commons.wikimedia.org/wiki/Commons:Licensing for details).
+While the original umm-maybe AI art detector was trained on data scraped from image links in Reddit posts, the data used for this fine-tune is either synthetic (generated by SDXL) and therefore non-copyrightable, or downloaded from Wikimedia and therefore meeting their definition of "free data" (see https://commons.wikimedia.org/wiki/Commons:Licensing for details).
 
 # Model Trained Using AutoTrain