OOD 泛化

[kirklandWater1]

感谢分享！请问有ood 泛化（eg: aime, math, etc.）的测试结果吗？

[Tim-Siu]

这个我没有测过。我也挺好奇的，不知道训了这么多step会不会overfit到这类题上了🫣

Logic-RL里好像上传了aime测试的script。理论上我们可以使用那个测试。

[kirklandWater1]

可以分享个hf上训练好的check point吗？我可以帮忙测一下

[Tim-Siu]

可以分享个hf上训练好的check point吗？我可以帮忙测一下

Hi, 我把checkpoint上传到modelscope了.

使用的时候可能要设置system message

INSTRUCT_PROMPT = """You are a helpful assistant. The assistant first thinks about the reasoning process in the mind and then provides the user with the answer. The reasoning process and answer are enclosed within <think> </think> and<answer> </answer> tags, respectively, i.e., <think> reasoning process here </think><answer> answer here </answer>.  Now the user asks you to solve a logical reasoning problem. After thinking, when you finally reach a conclusion, clearly state the identity of each character within <answer> </answer> tags. i.e., <answer> (1) Zoey is a knight\n(2) ... </answer>."""