This repository contains the implementation of token-level density-based methods for Uncertainty Quantification (UQ) for selective text generation tasks using LLMs. Namely, the repository contains ...
In this project, we reproduced the main results presented in [1]. Thus, we built from scratch (using PyTorch) the Mahalanobis-based score with Input pre-processing and feature ensembling as suggested ...
The Mahalanobis-Taguchi System (MTS) has evolved into a robust tool for multivariate data classification and optimisation, merging statistical distance measures with the design philosophy of robust ...
Abstract: In out-of-distribution detection, methods based on the Mahalanobis distance have excellent performance for detecting out-of-distribution data. The Mahalanobis distance measures the distance ...
The Mahalanobis distances have been introduced in habitat selection studies for the estimation of environmental suitability maps (ESMs). The pixels of raster maps of a given area correspond to points ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback