AI Paper Atlas

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

Richer Bayesian Last Layers with Subsampled NTK Features

Sergio Calvo Ordoñez, Jonathan Plenk, Richard Bergna, Alvaro Cartea, Yarin Gal, José Miguel Hernández-Lobato, Kamil Ciosek

ICML 2026 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Priors and functional variational uncertainty · 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
add NTK low-rank correction (Rich-BLL) to the Bayesian last layerNNGP (BLL) · UCI Concretetest NLL: 3.39 -> 3.10Improved
add NTK low-rank correction (Rich-BLL) to the Bayesian last layerNNGP (BLL) · UCI Bostontest NLL: 2.78 -> 2.61Improved
add NTK low-rank correction (Rich-BLL) to the Bayesian last layerNNGP (BLL) · UCI Energytest NLL: 0.92 -> 0.75Improved
add subsampling (40% of data) to Rich-BLL (ablation)Rich-BLL · UCI regressiontest NLL: 0.75 -> 0.78No clear effect
add NTK correction to the Bayesian last layerNNGP (BLL) · Wheel bandit delta=0.99normalized cumulative regret: 55.80 -> 21.80Improved
add subsampling to Rich-BLL (ablation)Rich-BLL · Wheel bandit delta=0.99normalized cumulative regret: 21.80 -> 22.30No clear effect
add NTK correction to the Bayesian last layerNNGP (BLL) · Wine Quality OODAUROC: 0.88 -> 0.96Improved
add NTK correction (subsampled) to the Bayesian last layerNNGP (BLL) · CIFAR-10NLL: 0.58 -> 0.56Improved
add NTK correction (subsampled) to the Bayesian last layerNNGP (BLL) · CIFAR-10 vs SVHNAUROC: 0.88 -> 0.91Improved
add NTK correction (subsampled) to the Bayesian last layerNNGP (BLL) · CIFAR-10 vs CIFAR-100AUROC: 0.62 -> 0.65Improved
use a single global transformation instead of class-specific ones (ablation)Rich-BLL classification · CIFAR-10accuracy/uncertaintyImproved

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