A new benchmark reveals that language models including GPT-3 are truthful on only 58% of questions designed to elicit popular misconceptions, far below human performance of 94%, with larger models performing worse.
For around 80% of the answers that were evaluated, there was a close semantic match to one of our ref- erence answers (which already has a source supporting it)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2021 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
TruthfulQA: Measuring How Models Mimic Human Falsehoods
A new benchmark reveals that language models including GPT-3 are truthful on only 58% of questions designed to elicit popular misconceptions, far below human performance of 94%, with larger models performing worse.