II.
Benchmark overview
Reference · livebenchmark:truthful-qa
TruthfulQA overview
TruthfulQA (Lin et al., 2021) is an 817-question benchmark measuring whether a language model produces truthful answers when humans tend to hold false beliefs (misconceptions, urban legends). Reports both truthfulness and informativeness scores.
Attributes
displayName
TruthfulQA
homepageUrl
kind
model-only
targetsKind
ModelVersion
description
TruthfulQA (Lin et al., 2021) is an 817-question benchmark
measuring whether a language model produces truthful answers
when humans tend to hold false beliefs (misconceptions, urban
legends). Reports both truthfulness and informativeness scores.
Outgoing edges
None.
Incoming edges
for_benchmark1
scored_against1
- eval-result:truthful-qa.claude-opus-4-5.001·EvalResult