Publication:
On measuring social biases in prompt-based multi-task learning

Thumbnail Image

School / College / Institute

Program

KU Authors

Co-Authors

Akyürek
A.F.
Paik, S.
Koçyiğit, M.Y.
Akbıyık, S.
Wijaya, D.

Publication Date

Language

Embargo Status

NO

Journal Title

Journal ISSN

Volume Title

Alternative Title

Abstract

Large language models trained on a mixture of NLP tasks that are converted into a textto- text format using prompts, can generalize into novel forms of language and handle novel tasks. A large body of work within prompt engineering attempts to understand the effects of input forms and prompts in achieving superior performance. We consider an alternative measure and inquire whether the way in which an input is encoded affects social biases promoted in outputs. In this paper, we study T0, a large-scale multi-task text-to-text language model trained using prompt-based learning. We consider two different forms of semantically equivalent inputs: question-answer format and premise-hypothesis format. We use an existing bias benchmark for the former BBQ (Parrish et al., 2021) and create the first bias benchmark in natural language inference BBNLI with hand-written hypotheses while also converting each benchmark into the other form. The results on two benchmarks suggest that given two different formulations of essentially the same input, T0 conspicuously acts more biased in question answering form, which is seen during training, compared to premisehypothesis form which is unlike its training examples.

Source

Publisher

Association for Computational Linguistics (ACL)

Subject

Psychology, Learning systems, Natural language processing systems

Citation

Has Part

Source

Findings of the Association for Computational Linguistics: NAACL 2022

Book Series Title

Edition

DOI

10.18653/v1/2022.findings-naacl.42

item.page.datauri

Link

Rights

Copyrights Note

Endorsement

Review

Supplemented By

Referenced By

1

Views

3

Downloads

View PlumX Details