LLM Neuroanatomy II: Modern LLM Hacking and Hints of a Universal Language?

2 comments

• vibe42 12 hours ago

  Pretty sweet hack as it's orthogonal to quantisation. And while it uses more compute, it doesn't require more VRAM.

  Maybe in the future circuits will become modular and composable like models are today?

• dnhkng 5 hours ago

  [dead]