What is a good score range for the val chart?

[yfen310]

Hi, thanks for the great work!

I’ve been training a model following your setup. I’d like to know: what’s a good range for val/test_score/JudgeLRM?

Just want to better understand how my results compare. Thanks in advance!

[NuoJohnChen]

Thank you for your attention to our work. Due to recent busyness, I’ve uploaded partial code for training to facilitate reproduction, and I plan to update the complete code asap. Please note that I have not modified ray_trainer.py in LogicRL or paid attention to changes in val/test_score/JudgeLRM. I mainly focused on the stable growth of the training set reward and changes in response_length.