You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm curious about the comparison results between prompt tuning and the current parameter-efficient tuning methods (e.g., Adapter, prefix-tuning, LoRA, BitFit, etc). Have you tried to compare with these models? Noting that in this preprint paper, the performance of prompt tuning (or SPoT) is less than Adapter (or BitFit). Is this because the number of trainable parameters by the prompt-tuning-style is too small?
Looking forward to hear your valuable view!
Thanks in advance.
Hi, thank you for sharing your wonderful work!
I'm curious about the comparison results between prompt tuning and the current parameter-efficient tuning methods (e.g., Adapter, prefix-tuning, LoRA, BitFit, etc). Have you tried to compare with these models? Noting that in this preprint paper, the performance of prompt tuning (or SPoT) is less than Adapter (or BitFit). Is this because the number of trainable parameters by the prompt-tuning-style is too small?
Looking forward to hear your valuable view!
Thanks in advance.