"Discovered a very interesting thing about DeepSeek-R1 and all reasoning models: The wrong answers are much longer while the correct answers are much shorter."
svenfaw
https://hexatomium.github.io/
https://twitter.com/hexatomsPosts258
Comments76