Safety, alignment & interpretability· 2021
TruthfulQA: Measuring How Models Mimic Human Falsehoods
DOI 10.18653/v1/2022.acl-long.229arXiv 2109.07958OpenAlex W3201174429
On the reading lists
- #433 in the Top 500 (2026 edition)
- #45 in Safety, alignment & interpretability (2026 edition)
Why it matters
Introduced a benchmark exposing language models' tendency to reproduce common human misconceptions.
Condensed from the expert lists' notes.
The record
- First released
- 2021 (from arXiv 2109.07958)
- Citations
- 4,109 citations, as counted by Papers With Code (CC BY-SA 4.0)
- Named by
- both expert lists
- Score
- 0.368 · computed rank 433 · consensus 0.21, must-read 0, citations percentile 0.872 (2020s · safety), awards 0, existing lists 0. Method
Data sources
Catalogue data from OpenAlex (CC0), Crossref (CC0 metadata; no abstracts are shown), arXiv, and Papers With Code (CC BY-SA 4.0). Each count names its source. See the method.