TruthfulQA: Measuring how models mimic human falsehoods

TruthfulQA benchmark measures language model propensity to repeat human falsehoods, becoming standard evaluation for truthfulness.

Categories: Research

Read at source: https://openai.com/index/truthfulqa