Fixing GRPO's credit assignment problem without evaluating every step

[mrkn1]

https://arxiv.org/abs/2609.36178

[vatsachak]

How do frontier labs choose which ideas to include in a training run? It seems that there are too many to choose from

[martianvoid → vatsachak]

I think maybe they have enough compute to try out all the promising directions at a smaller scale first and then port them to larger run

They also have incredibly smart people that can understand which improvements and directions are even worth pursuing for, I think these people are being paid in millions

[simianwords → vatsachak]

Why do you think they are buying compute and accelerating to RSI? It’s for this.