LLMs and Computation Complexity (2023)

(lesswrong.com)

3 points | by mathattack 13 hours ago

1 comments

  • tomeraberbach 12 hours ago
    Question: did the LLM need O(n^3) tokens to solve the matrix multiplication example? Or would O(n^1.5) tokens be enough? Since producing that many tokens is O(n^3) work if I understand currently?