AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

Testing Most Influential Sets

Lucas Darius Konrad, Nikolas Kuschnig

ICLR 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Adaptive testing and experiments

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
test the Seychelles' influence with the EVT hypothesis testruggedness regression · Ruggedness vs GDPp-value: p < 0.001 aloneImproved
pair Seychelles with Lesothoruggedness regression · Ruggedness vs GDPp-value: p = 0.216No clear effect
test the 77-point set that flips the sign vs the 17-point setLaw School regression · Law Schoolexcessive influence: 17 points p = 0.019; 77 points within expected variationMixed
test top 1 percent influential sets (325 points)Adult Income regression · Adult Incomeexcessive influence: 0.062 -> 0.0992 or 0.0214No clear effect
use Frechet (heavy-tailed crime variable) null for 6-point setBoston Housing regression · Boston Housingp-value: -0.1080 -> -0.0352; p = 0.001Improved
bias-correct the Gumbel location estimate in block maxima (ablation)EVD parameter estimation · simulationlocation estimateImproved
heavy-tailed X (t(5)) with normal residuals instead of normal-normal (ablation)EVD convergence · simulation N=20 to 1000convergence speedHurt
use exact set influence formula with leave-out Gram matrix instead of influence function approximationinfluence of sets · theoryaccuracy of influenceImproved

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