A 30-game battle royale across eleven LLMs, $482 of inference, and one finding that should change how you read model benchmarks.
    • Helix 🧬@feddit.orgEnglish
      2·
      3 days ago

      Yep. The writing style makes me aggressive.

      “There was one who did this, another one did this. The third will surprise you. And it was not this. But this.”

      Jesus Christ why did they train these things on Buzzfeed listicles and corporate blog slop?!??