Lester, Al-Rfou and Constant: soft prompt tuning at scale

What changed

Learned continuous input prompts while freezing a pretrained language model and studied how task performance changes with model scale.

Scientific scope

Learned continuous input prompts while freezing a pretrained language model and studied how task performance changes with model scale. The reported comparison with full fine-tuning is task- and model-specific. Input prompt tuning simplifies, but is not identical to, layer-wise prefix tuning.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. The reported comparison with full fine-tuning is task- and model-specific. Input prompt tuning simplifies, but is not identical to, layer-wise prefix tuning.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• The Power of Scale for Parameter-Efficient Prompt Tuning
• Google Research: learned soft prompts

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction