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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

We propose a method to control the attributes of Language Models (LMs) forthe text generation task using Causal Average Treatment Effect (ATE) scores andcounterfactual augmentation. We explore this method, in the context of LMdetoxification, and propose the Causally Fair Language (CFL) architecture fordetoxifying pre-trained LMs in a plug-and-play manner. Our architecture isbased on a Structural Causal Model (SCM) that is mathematically transparent andcomputationally efficient as compared with many existing detoxificationtechniques. We also propose several new metrics that aim to better understandthe behaviour of LMs in the context of toxic text generation. Further, weachieve state of the art performance for toxic degeneration, which are computedusing \RTP (RTP) benchmark. Our experiments show that CFL achieves such adetoxification without much impact on the model perplexity. We also show thatCFL mitigates the unintended bias problem through experiments on the BOLDdataset.

CFL: Causally Fair Language Models Through Token-level Attribute Controlled Generation