Evidential Transformation Network: Turning Pretrained Models into Evidential Models for Post-hoc Uncertainty Estimation

arXiv cs.AI / 4/13/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper notes that standard pretrained vision/language models often lack reliable, deployable confidence estimates, and that common uncertainty methods like deep ensembles and MC dropout can be too expensive in practice.
It proposes the Evidential Transformation Network (ETN), a lightweight post-hoc module that converts an existing pretrained predictor into an evidential model without retraining the base network from scratch.
ETN learns a sample-dependent affine transformation of logits and treats the transformed outputs as Dirichlet distribution parameters to produce evidential uncertainty estimates.
Experiments on image classification and LLM question-answering benchmarks show ETN improves uncertainty estimation in both in-distribution and out-of-distribution settings while preserving predictive accuracy and adding minimal computational overhead.

Abstract

Pretrained models have become standard in both vision and language, yet they typically do not provide reliable measures of confidence. Existing uncertainty estimation methods, such as deep ensembles and MC dropout, are often too computationally expensive to deploy in practice. Evidential Deep Learning (EDL) offers a more efficient alternative, but it requires models to be trained to output evidential quantities from the start, which is rarely true for pretrained networks. To enable EDL-style uncertainty estimation in pretrained models, we propose the Evidential Transformation Network (ETN), a lightweight post-hoc module that converts a pretrained predictor into an evidential model. ETN operates in logit space: it learns a sample-dependent affine transformation of the logits and interprets the transformed outputs as parameters of a Dirichlet distribution for uncertainty estimation. We evaluate ETN on image classification and large language model question-answering benchmarks under both in-distribution and out-of-distribution settings. ETN consistently improves uncertainty estimation over post-hoc baselines while preserving accuracy and adding only minimal computational overhead.