The Sensitivity of Language Models and Humans to Winograd Schema Perturbations
Publikation: Bidrag til bog/antologi/rapport › Konferencebidrag i proceedings › Forskning › fagfællebedømt
Dokumenter
- The Sensitivity of Language Models and Humans
Forlagets udgivne version, 633 KB, PDF-dokument
Large-scale pretrained language models are the major driving force behind recent improvements in perfromance on the Winograd Schema Challenge, a widely employed test of commonsense reasoning ability. We show, however, with a new diagnostic dataset, that these models are sensitive to linguistic perturbations of the Winograd examples that minimally affect human understanding. Our results highlight interesting differences between humans and language models: language models are more sensitive to number or gender alternations and synonym replacements than humans, and humans are more stable and consistent in their predictions, maintain a much higher absolute performance, and perform better on non-associative instances than associative ones.
Originalsprog | Engelsk |
---|---|
Titel | Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics |
Forlag | Association for Computational Linguistics |
Publikationsdato | 2020 |
Sider | 7590-7604 |
DOI | |
Status | Udgivet - 2020 |
Begivenhed | 58th Annual Meeting of the Association for Computational Linguistics - Online Varighed: 5 jul. 2020 → 10 jul. 2020 |
Konference
Konference | 58th Annual Meeting of the Association for Computational Linguistics |
---|---|
By | Online |
Periode | 05/07/2020 → 10/07/2020 |
Antal downloads er baseret på statistik fra Google Scholar og www.ku.dk
Ingen data tilgængelig
ID: 258374819