English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Despite the success, the process of fine-tuning large-scale PLMs bringsprohibitive adaptation costs. In fact, fine-tuning all the parameters of acolossal model and retaining separate instances for different tasks arepractically infeasible. This necessitates a new branch of research focusing onthe parameter-efficient adaptation of PLMs, dubbed as delta tuning in thispaper. In contrast with the standard fine-tuning, delta tuning only fine-tunesa small portion of the model parameters while keeping the rest untouched,largely reducing both the computation and storage costs. Recent studies havedemonstrated that a series of delta tuning methods with distinct tunedparameter selection could achieve performance on a par with full-parameterfine-tuning, suggesting a new promising way of stimulating large-scale PLMs. Inthis paper, we first formally describe the problem of delta tuning and thencomprehensively review recent delta tuning approaches. We also propose aunified categorization criterion that divide existing delta tuning methods intothree groups: addition-based, specification-based, and reparameterization-basedmethods. Though initially proposed as an efficient method to steer largemodels, we believe that some of the fascinating evidence discovered along withdelta tuning could help further reveal the mechanisms of PLMs and even deepneural networks. To this end, we discuss the theoretical principles underlyingthe effectiveness of delta tuning and propose frameworks to interpret deltatuning from the perspective of optimization and optimal control, respectively.Furthermore, we provide a holistic empirical study of representative methods,where results on over 100 NLP tasks demonstrate a comprehensive performancecomparison of different approaches. The experimental results also cover theanalysis of combinatorial, scaling and transferable properties of delta tuning.

Delta Tuning: A Comprehensive Study of Parameter Efficient Methods for Pre-trained Language Models