AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Is the Last Layer Sufficient for Uncertainty Quantification?

Joseph Wilson, Chris van der Heide, Liam Hodgkinson, Fred Roosta

ICML 2026 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Epistemic versus aleatoric decomposition · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
linearize only the last layer (LL-GLM) instead of the full network (DNN-GLM)Bayesian GLM UQ · UCI regression (9 datasets)RMSE, NLL, ECENo clear effect
use last-layer GLM instead of full GLMGPT-2 fine-tuned on IMDB · IMDBVARROC-ID: 0.842 -> 0.823No clear effect
use last-layer GLM instead of full GLMGPT-2 fine-tuned on IMDB · IMDBwall-time and memory: 2.55 hr -> 0.26 hr; 71.23 GB -> 20.97 GBImproved
use last-layer GLM instead of full GLMimage classifiers (LeNet, ResNet9, ResNet50) · image classification, 8 model/dataset pairsVARROC and LPPD: average relative difference V-ID -1.14%, V-OOD 1.20%No clear effect
use last-layer GLM instead of full GLMimage classifiers · image classification, 8 model/dataset pairstime and memory: average reduction 58.44% time, 57.32% memoryImproved
use full NTK (DNN-GLM) instead of CK (LL-GLM) at random initialization when training points greatly exceed widthGP kernel at initialization · theory, over-sampled regimeBayes free energyImproved
train the network before forming the CK and NTKCK vs NTK · teacher-network and Gaussian dataBayes free energyImproved
restrict weak descent to NTK versus CK parameter regionCK vs NTK · theory, weak descentdescent regionMixed
compare LL-GLM and DNN-GLM against deep ensemblesGLMs · image classificationVARROCHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy