ConfidenceBench scores 15 frontier LLMs on how well their stated confidence matches reality, and finds accuracy and calibration often don't track together.
Continue to AI University →