3 ms·

This seems like a near perfect use of coding LLMs and a useful way to implement reinforcement learning. “Add a major bug to this file that is not covered by ex

• infogulch 2y ago
  Somehow that reminds me of how diffusion models are trained.
  • _lvbh 2y ago
    I'm pretty sure that's the premise for GANs (generative adversarial networks) rather than diffusion. Diffusion is more about noise reduction than pitting models against each other