Matches in SemOpenAlex for { <https://semopenalex.org/work/W4307206531> ?p ?o ?g. }
Showing items 1 to 65 of
65
with 100 items per page.
- W4307206531 abstract "Derivative-free prompt learning has emerged as a lightweight alternative to prompt tuning, which only requires model inference to optimize the prompts. However, existing work did not take full advantage of the over-parameterized characteristics of large pre-trained language models (PLMs). In this paper, we propose Clip-Tuning, a simple yet effective method that adopts diverse frozen thinned networks of PLMs to obtain a mixture of rewards and thus advance the derivative-free prompt learning. The thinned networks consist of all the hidden units that survive a stationary dropout strategy, whose inference predictions reflect an ensemble of partial views over prompted training samples. Our method outperforms previous gradient-free prompt learning methods and achieves parity with gradient-based counterparts on seven language understanding benchmarks under few-shot settings." @default.
- W4307206531 created "2022-10-30" @default.
- W4307206531 creator A5023307885 @default.
- W4307206531 creator A5036350345 @default.
- W4307206531 creator A5043069455 @default.
- W4307206531 creator A5058880150 @default.
- W4307206531 creator A5072082849 @default.
- W4307206531 creator A5084573823 @default.
- W4307206531 date "2022-10-21" @default.
- W4307206531 modified "2023-09-25" @default.
- W4307206531 title "Clip-Tuning: Towards Derivative-free Prompt Learning with a Mixture of Rewards" @default.
- W4307206531 doi "https://doi.org/10.48550/arxiv.2210.12050" @default.
- W4307206531 hasPublicationYear "2022" @default.
- W4307206531 type Work @default.
- W4307206531 citedByCount "0" @default.
- W4307206531 crossrefType "posted-content" @default.
- W4307206531 hasAuthorship W4307206531A5023307885 @default.
- W4307206531 hasAuthorship W4307206531A5036350345 @default.
- W4307206531 hasAuthorship W4307206531A5043069455 @default.
- W4307206531 hasAuthorship W4307206531A5058880150 @default.
- W4307206531 hasAuthorship W4307206531A5072082849 @default.
- W4307206531 hasAuthorship W4307206531A5084573823 @default.
- W4307206531 hasBestOaLocation W43072065311 @default.
- W4307206531 hasConcept C106159729 @default.
- W4307206531 hasConcept C111472728 @default.
- W4307206531 hasConcept C111771559 @default.
- W4307206531 hasConcept C11413529 @default.
- W4307206531 hasConcept C119857082 @default.
- W4307206531 hasConcept C138885662 @default.
- W4307206531 hasConcept C154945302 @default.
- W4307206531 hasConcept C162324750 @default.
- W4307206531 hasConcept C165464430 @default.
- W4307206531 hasConcept C2776145597 @default.
- W4307206531 hasConcept C2776214188 @default.
- W4307206531 hasConcept C2780586882 @default.
- W4307206531 hasConcept C41008148 @default.
- W4307206531 hasConceptScore W4307206531C106159729 @default.
- W4307206531 hasConceptScore W4307206531C111472728 @default.
- W4307206531 hasConceptScore W4307206531C111771559 @default.
- W4307206531 hasConceptScore W4307206531C11413529 @default.
- W4307206531 hasConceptScore W4307206531C119857082 @default.
- W4307206531 hasConceptScore W4307206531C138885662 @default.
- W4307206531 hasConceptScore W4307206531C154945302 @default.
- W4307206531 hasConceptScore W4307206531C162324750 @default.
- W4307206531 hasConceptScore W4307206531C165464430 @default.
- W4307206531 hasConceptScore W4307206531C2776145597 @default.
- W4307206531 hasConceptScore W4307206531C2776214188 @default.
- W4307206531 hasConceptScore W4307206531C2780586882 @default.
- W4307206531 hasConceptScore W4307206531C41008148 @default.
- W4307206531 hasLocation W43072065311 @default.
- W4307206531 hasOpenAccess W4307206531 @default.
- W4307206531 hasPrimaryLocation W43072065311 @default.
- W4307206531 hasRelatedWork W2100045227 @default.
- W4307206531 hasRelatedWork W2605524926 @default.
- W4307206531 hasRelatedWork W2953328427 @default.
- W4307206531 hasRelatedWork W2963058055 @default.
- W4307206531 hasRelatedWork W3042605146 @default.
- W4307206531 hasRelatedWork W3186919929 @default.
- W4307206531 hasRelatedWork W4210794429 @default.
- W4307206531 hasRelatedWork W4280592718 @default.
- W4307206531 hasRelatedWork W4307206531 @default.
- W4307206531 hasRelatedWork W4312863455 @default.
- W4307206531 isParatext "false" @default.
- W4307206531 isRetracted "false" @default.
- W4307206531 workType "article" @default.