AI Paper Atlas

Research map / Probabilistic and scientific modeling / Bayesian optimization

HyperSHAP: Shapley Values and Interactions for Explaining Hyperparameter Optimization

Marcel Wever, Maximilian Muschalik, Fabian Fumagalli, Marius Lindauer

AAAI 2026 · Probabilistic and scientific modeling › Bayesian optimization · approach: Offline black-box optimization with surrogates · 3 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
use Tunability game instead of Ablation game for explaining an optimized configuration (ablation)HyperSHAP · PD1 lm1b transformerexplanation of hyperparameter importanceImproved
select top-2 hyperparameters via HyperSHAP instead of fANOVA before HPOHPO run on two hyperparameters · lcbench (two datasets)anytime performanceImproved
use Tunability game instead of Sensitivity game to pick hyperparameters to tune (ablation)HyperSHAP · lcbench (two datasets)anytime performanceImproved
raise Shapley interaction explanation order up to 3 (ablation)HyperSHAP · HPO benchmarksfaithfulness (R2): R2 ≈ 1 at order 3Improved
raise explanation order beyond third (ablation)HyperSHAP · HPO benchmarksfaithfulness (R2)No clear effect
use Optimizer Bias game to detect optimizer that tunes hyperparameters independentlyHyperSHAP · lcbench dataset 3945bias detectionImproved
use Optimizer Bias game to detect optimizer that skips weight decay tuningHyperSHAP · lcbench dataset 3945bias detectionImproved

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