Uncertainty Estimation
A technique in deep learning for quantifying the confidence of model predictions. It is divided into two categories, epistemic uncertainty and aleatoric uncertainty, and common implementations include Monte Carlo Dropout, deep ensembles, and Bayesian neural networks.
Timeline (last 90 days)
Existing uncertainty estimation schemes usually require multiple samplings (5x or more compute), the introduction of a judge model, or reliance on labeled datasets, making them costly and only giving general confidence.
Treating confidence as a proxy for knowledge is the implicit premise of many evaluation, hallucination detection, and uncertainty estimation methods.
All Facts (2)
Existing uncertainty estimation schemes usually require multiple samplings (5x or more compute), the introduction of a judge model, or reliance on labeled datasets, making them costly and only giving general confidence.
50%UnverifiedTreating confidence as a proxy for knowledge is the implicit premise of many evaluation, hallucination detection, and uncertainty estimation methods.
50%