Missing References in the Prompt Tuning section.

[MLWatcher]

There is a large body of prior and concurrent work in prompt tuning that has been left out of this paper. For example:

• Learning How to Ask, one of the first to learn a continuous version of a prompt https://arxiv.org/abs/2104.06599
• Prefix-Tuning https://arxiv.org/abs/2101.00190
• Prompt Tuning, where you seem to have gotten the term model-tuning from https://arxiv.org/abs/2104.08691
• GPT Understands Too, which jointly learns a prompt and updates the model like the second step of your two step pipeline https://arxiv.org/abs/2103.10385
• WARP https://arxiv.org/abs/2101.00121

[Zhengyan Zhang]

Hi,

Thanks for your suggestion. We have added these references to our latest version.