update layout

Author: simonucl

app/blog/layout.tsx

@@ -1,26 +1,26 @@
 import type { Metadata } from 'next';
 
 export const metadata: Metadata = {
-  title: 'Verbalized Sampling Blog',
-  description: 'Verbalized Sampling: A training-free method to restore diversity in aligned LLMs by prompting for probability distributions.',
+  title: 'Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity | Blog',
+  description: 'Blog showcasing Verbalized Sampling—a training-free method that restores diversity in aligned LLMs by 1.6-2.1× by asking for probability distributions instead of single answers. Explore results, examples, and try it yourself.',
   openGraph: {
-    title: 'Verbalized Sampling Blog',
-    description: 'Verbalized Sampling: A training-free method to restore diversity in aligned LLMs by prompting for probability distributions.',
+    title: 'Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity | Blog',
+    description: 'Blog showcasing Verbalized Sampling—a training-free method that restores diversity in aligned LLMs by 1.6-2.1× by asking for probability distributions instead of single answers. Explore results, examples, and try it yourself.',
     url: 'https://www.verbalized-sampling.com/blog/',
     type: 'website',
     images: [
       {
         url: 'https://www.verbalized-sampling.com/og-qualitative.jpg',
         width: 1200,
         height: 630,
-        alt: 'Verbalized Sampling Blog'
+        alt: 'Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity | Blog'
       }
     ]
   },
   twitter: {
     card: 'summary_large_image',
-    title: 'Verbalized Sampling Blog',
-    description: 'Verbalized Sampling: A training-free method to restore diversity in aligned LLMs by prompting for probability distributions.',
+    title: 'Verbalized Sampling: How to Mitigate Mode Collapse and Unlock LLM Diversity | Blog',
+    description: 'Blog showcasing Verbalized Sampling—a training-free method that restores diversity in aligned LLMs by 1.6-2.1× by asking for probability distributions instead of single answers. Explore results, examples, and try it yourself.',
     images: ['https://www.verbalized-sampling.com/og-qualitative.jpg']
   }
 };