🐛 Add missing </p>

Author: o-laurent

index.html

@@ -204,6 +204,7 @@ <h2 class="title is-3">Abstract</h2>
                 LS degrades the uncertainty rank ordering of correct vs incorrect predictions by regularising the max logit more when a prediction is likely to be correct, and less when it is likely to be wrong. 
                 This elucidates previously reported experimental results where strong classifiers underperform in SC. We then demonstrate the empirical effectiveness of post-hoc logit normalisation for recovering lost SC performance caused by LS. 
                 Furthermore, linking back to our gradient analysis, we again provide an explanation for why such normalisation is effective. <br><br>
+            </p>
           </div>
         </div>
       </div>