About the experimental results of RAP

Open zy0055 opened this issue 2 years ago • 3 comments

Hello, I used RAP for defense on sst-2, but the experimental result was very poor, the FAR was close to 1.0, may I ask if it is normal? If it is not normal, how should I set the details of the experiment

Jul 16 '23 13:07 zy0055

May i ask whether your problem is solved yet? i met the same problem when i run demo_defend.py too.

Apr 24 '24 07:04 Vinc3ent

May i ask whether your problem is solved yet? i met the same problem when i run demo_defend.py too.

Finally, I used the code of the original paper and modified the learning rate

Apr 24 '24 08:04 zy0055

Thanks a lot. Since I have some further questions to consult, could we possibly communicate on Wechat, please? If it won't bother you, this is my Wechat account: Vir_cent0915. Thanks again.

Apr 24 '24 10:04 Vinc3ent