g1: Using Llama-3.1 70B on Groq to create o1-like reasoning chains

Re:g1: Using Llama-3.1 70B on Groq to create o1-like reasoni... by ricardobeat on Monday September 16, 2024 @01:18PM (#41558265)

Care to elaborate? Your comment would be a lot more useful if it included a little why. Otherwise it’s just teasing readers and at the same time smearing the author without anything to back it up.

[ Parent | on HN ]

Re:g1: Using Llama-3.1 70B on Groq to create o1-like reasoni... by whimsicalism on Monday September 16, 2024 @02:28PM (#41559007)

reinforcement learning with ppo doesn’t involve mcts and has been the bread and butter of aligning LLMs since 2020. nothing about saying they use rl implies mcts

[ Parent | on HN ]

Re:g1: Using Llama-3.1 70B on Groq to create o1-like reasoni... by janalsncm on Monday September 16, 2024 @04:54PM (#41560792)

> nothing about saying they use rl implies they use mcts

We can say the same thing about RL implying PPO, however there’s pretty big hints, namely Noam Brown being involved. Many of the things Noam Brown has worked on involve RL in tree search contexts.

He has also been consistently advocating the use of additional test-time compute to solve search problems. This is also consistent with the messaging regarding the reasoning tokens. There is likely some learned tree search algorithm, such as a learned policy/value function as in AlphaGo.

It’s all speculation until we have an actual paper. So we can’t categorically say MCTS/learned tree search isn’t involved.

[ Parent | on HN ]