TruthfulQA: Measuring How Models Mimic Human Falsehoods
Introduces TruthfulQA, 817 adversarial questions designed so that imitating common human misconceptions yields wrong answers; larger models often do worse.
- Publisher
- University of Oxford
- Year
- 2021
- Venue
- ACL
- Authors
- 3
- Hosting
- External sourcelicense unknown
Cite
Notes
Only stored in your browser.
Introduces 1 artifact - 1 eval
TL;DR
Semantic Scholar
It is suggested that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web.
Artifacts
1Evals