ConfidenceBench: Evaluating Confidence Calibration in Large Language Models
For practitioners deploying LLMs in high-stakes settings, this benchmark provides a standardized way to evaluate a distinct reliability dimension—calibration—that is not captured by accuracy alone.
ConfidenceBench evaluates verbalized confidence calibration in 15 frontier LLMs using the Brier score, finding that the best-calibrated models (Claude Opus 4.6 and Gemini 3.1 Pro Preview, both with Brier score 0.103) substantially outperform the calibrated-random baseline (0.1875), while others like Gemini 3.1 Flash-Lite (0.367) show severe miscalibration, and accuracy does not correlate with calibration.
Large language models (LLMs) are increasingly deployed in settings where fluent but incorrect answers can be costly. In these settings, accuracy alone is insufficient: models must also know when they are likely to be wrong. We present ConfidenceBench, a calibration benchmark that evaluates verbalized confidence estimates in 15 frontier LLMs using the Brier score, a proper scoring rule that incentivises truthful probability reporting. Confidence is elicited via prompting, requiring no access to model logits and making the framework applicable to both closed-source and open-source systems. The benchmark comprises 200 private multiple-choice questions across four categories: spatial reasoning, high-precision mathematics, word lookup, and unknowable questions. Across three independent runs, Claude Opus 4.6 and Gemini 3.1 Pro Preview achieve the lowest Brier scores, both reported as 0.103. Both substantially outperform the calibrated-random baseline of 0.1875, while Gemini 3.1 Flash-Lite scores 0.367, indicating severe miscalibration. Accuracy and calibration diverge substantially across model families: the most accurate model is not the best-calibrated, and several models perform worse than the calibrated-random Brier baseline despite reasonable accuracy. These results show that verbalized confidence calibration is a distinct and practically important axis of LLM reliability, complementary to standard accuracy-based evaluation.