Home AI Hacker News Show HN: Analyzing Writing Styles and Similarity Clusters of 178 AI Models

Show HN: Analyzing Writing Styles and Similarity Clusters of 178 AI Models

0

Unlocking the Secrets of AI Response Analysis 🚀

Dive into our groundbreaking dataset of 3,095 standardized AI responses across 43 prompts, revealing the intricacies behind AI writing styles. Our stylometric analysis incorporates a 32-dimension fingerprint capturing essential traits like lexical richness and discourse markers.

Key Insights:

  • Clone Clusters: Identified 9 clusters with over 90% cosine similarity.
  • High Performers: Mistral Large 2 and Large 3 achieved an 84.8% score on a composite metric.
  • Cost-Effective: Gemini 2.5 Flash Lite writes 78% like Claude 3 Opus, at 185x less cost.
  • Distinctive Style: Meta showcases the strongest house style with a 37.5x distinctiveness ratio.

The prompt “Satirical fake news” demonstrates significant convergence, while “Count letters” triggers divergence.

Are you intrigued by AI’s evolving landscape? Share your thoughts or connect to explore further! 🧠✨

Source link

NO COMMENTS

Exit mobile version