I fine-tuned a 14B model to beat GPT-4o at NYT Connections (30% vs. 22.7%)(john463212.substack.com)
1 pointby Johnene2 hours ago1 comment
  • 2 hours ago
    undefined