AI Paper Atlas

Research map / Learning theory and algorithms / Kernels and regression theory

On the Bayes Inconsistency of Disagreement Discrepancy Surrogates

Neil G Marchant, Andrew Craig Cullen, Feng Liu, Sarah Monazam Erfani

ICLR 2026 · Learning theory and algorithms › Kernels and regression theory · approach: Statistical perspectives on generalization · 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
replace existing disagreement discrepancy surrogates (RG23, GLK23) with the proposed consistent surrogateerror-bound disagreement discrepancy maximization · 130 shift/model combinations, 11 vision benchmarksshare of instances reaching maximum disagreement discrepancy: almost 80% of instances reach the maximum; Wilcoxon p = 1.8 × 10^-11Improved
use the proposed surrogate instead of RG23 and GLK23 under adversarially chosen target dataerror-bound disagreement discrepancy maximization · Adversarial target data, f from 0% to 50%share of instances with highest estimate: 87.5% for f = 0% -> 100% for f = 50%Improved
use the proposed surrogate in the Detectron harmful-shift test instead of GLK23Detectron shift detection with XGBoost critics · UCI Heart Disease, 5 classes, N in {10,20,50}AUC: 95% confidence intervals for the AUC do not overlap for any NImproved
use the proposed surrogate for error-bound calibration instead of RG23'serror bound calibration · 11 vision benchmarkscalibration of error bounds across deltaMixed

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