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

Fine-tuning pretrained language models (LMs) without making any architecturalchanges has become a norm for learning various language downstream tasks.However, for non-language downstream tasks, a common practice is to employtask-specific designs for input, output layers, and loss functions. Forinstance, it is possible to fine-tune an LM into an MNIST classifier byreplacing the word embedding layer with an image patch embedding layer, theword token output layer with a 10-way output layer, and the word predictionloss with a 10-way classification loss, respectively. A natural questionarises: Can LM fine-tuning solve non-language downstream tasks without changingthe model architecture or loss function? To answer this, we proposeLanguage-Interfaced Fine-Tuning (LIFT) and study its efficacy and limitationsby conducting an extensive empirical study on a suite of non-languageclassification and regression tasks. LIFT does not make any changes to themodel architecture or loss function, and it solely relies on the naturallanguage interface, enabling "no-code machine learning with LMs." We find thatLIFT performs comparably well across a wide range of low-dimensionalclassification and regression tasks, matching the performances of the bestbaselines in many cases, especially for the classification tasks. We alsoreport experimental results on the fundamental properties of LIFT, includinginductive bias, robustness, and sample complexity. We also analyze the effectof pretraining on LIFT and a few properties/techniques specific to LIFT, e.g.,context-aware learning via appropriate prompting, calibrated predictions, datageneration, and two-stage fine-tuning. Our code is available athttps://github.com/UW-Madison-Lee-Lab/LanguageInterfacedFineTuning.

LIFT: Language-Interfaced Fine-Tuning for Non-Language Machine Learning Tasks