Generalization to larger models, eg. Qwen 3.8-27B

[cktang88]

In the docs/examples I see only pretty small models used, is there any evidence that say, Qwen 3.8-27B would benefit from this?

[hzy46]

It is standard reinforcement learning thus Qwen 3.8-27B will definitely benefit.

But the problem is train Qwen 3.8-27B will need more computing resource.