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

Current pre-trained language models have enabled remarkable improvements indownstream tasks, but it remains difficult to distinguish effects ofstatistical correlation from more systematic logical reasoning grounded on theunderstanding of real world. We tease these factors apart by leveragingcounterfactual conditionals, which force language models to predict unusualconsequences based on hypothetical propositions. We introduce a set of testsfrom psycholinguistic experiments, as well as larger-scale controlled datasets,to probe counterfactual predictions from five pre-trained language models. Wefind that models are consistently able to override real-world knowledge incounterfactual scenarios, and that this effect is more robust in case ofstronger baseline world knowledge -- however, we also find that for most modelsthis effect appears largely to be driven by simple lexical cues. When wemitigate effects of both world knowledge and lexical cues to test knowledge oflinguistic nuances of counterfactuals, we find that only GPT-3 showssensitivity to these nuances, though this sensitivity is also non-triviallyimpacted by lexical associative factors.

Counterfactual reasoning: Testing language models' understanding of hypothetical scenarios