natufunu

3 ms·

1. 1.▲
   by natufunu 7mo ago

   Great post! I wonder why MCTS is not more popular as a test time compute harness. Did you compare performance of MCTS (without distillation) against other methods (eg best of N) with the same compute budget?
2. 2.▲
   by natufunu 9mo ago

   One thing I learned from this is that my prompts are much less detailed than what author has been using. Very cool work and great write up.
3. 3.▲
   by natufunu 5y ago

   I wrote a hex AI with a couple of friends a few years back based on reinforcement learning. https://cleeff.github.io/hex/ You move first and the ai might swap meaning that it steals your first move and you swap colors.
4. 4.▲
   Show HN: Play Hex against our RL agent(cleeff.github.io)

   2 points by natufunu 6y ago|0 comments