Taeyup Kim

have you compared your GRPO model with a fine-tuned model?
On deepseek paper, they say that finetuning+GRPO is much more effective for warming up the models.
Also, I think evaluation between GRPO vs Fine-tuning could be meaningful.