uncertainty estimates
Uncertainty estimates in AI refer to the quantification of confidence in predictions made by models. This can involve measuring epistemic uncertainty (uncertainty in the model itself) and aleatoric uncertainty (uncertainty inherent in the data). They are important for applications requiring reliability, such as medical diagnosis or autonomous driving.
- AcuRank: Uncertainty-Aware Adaptive Computation for Listwise Reranking
- Efficient semantic uncertainty quantification in language models via diversity-steered sampling
- NeuralSurv: Deep Survival Analysis with Bayesian Uncertainty Quantification
- Quantifying Uncertainty in the Presence of Distribution Shifts
- Robust Sampling for Active Statistical Inference
- Robust and Computation-Aware Gaussian Processes
- Semi-Supervised Regression with Heteroscedastic Pseudo-Labels
- Test Time Scaling for Neural Processes
- Torch-Uncertainty: Deep Learning Uncertainty Quantification
- Uncertainty Estimation on Graphs with Structure Informed Stochastic Partial Differential Equations