Upload folder using huggingface_hub

hate_speech_demo.py

@@ -789,16 +789,17 @@ def create_gradio_app():
     </ul>
 
     <h2>How it works</h2>
-    <ul style="line-height: 1.7;">
-        <li><strong>Document-Grounded Evaluations</strong>: Every rating is directly tied to our 
-            <a href="#" onclick="openPolicyPopup(); return false;">
-                hate speech policy document
-            </a>, which makes our system far superior to other solutions that lack transparent decision criteria.
-        </li>
-        <li><strong>Adaptable Policies</strong>: The policy document serves as a starting point and can be easily adjusted to meet your specific requirements. As policies evolve, the system immediately adapts without requiring retraining.</li>
-        <li><strong>Clear Rationales</strong>: Each evaluation includes a detailed explanation referencing specific policy sections, allowing users to understand exactly why content was flagged or approved.</li>
-        <li><strong>Continuous Improvement</strong>: The system learns from feedback, addressing any misclassifications by improving retrieval accuracy over time.</li>
-    </ul>
+    <p><strong>Document-Grounded Evaluations</strong>: Every rating is directly tied to our 
+    <a href="#" onclick="openPolicyPopup(); return false;">
+        hate speech policy document
+    </a>, which makes our system far superior to other solutions that lack transparent decision criteria.
+    </p>
+
+<p><strong>Adaptable Policies</strong> The policy document serves as a starting point and can be easily adjusted to meet your specific requirements. As policies evolve, the system immediately adapts without requiring retraining.</p>
+
+<p><strong>Clear Rationales</strong> Each evaluation includes a detailed explanation referencing specific policy sections, allowing users to understand exactly why content was flagged or approved.</p>
+
+<p><strong>Continuous Improvement</strong> The system learns from feedback, addressing any misclassifications by improving retrieval accuracy over time.</p>   
 
     <p>Our approach combines Contextual's state-of-the-art 
         <a href='https://contextual.ai/blog/introducing-instruction-following-reranker/' target='_blank'>steerable reranker</a>,