AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

TabPack: Efficient Hyperparameter Ensembles for Tabular Deep Learning

Yury Gorishniy, Akim Kotelnikov, Ivan Rubachev, Artem Babenko

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Dataset condensation and flow estimation · 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
tune a shared set of hyperparameters per dataset for all base models instead of random sampling (SameHP) (ablation)TabPack† · 17 tabular datasetsmean rank: 1.6 ± 0.6 -> 1.4 ± 0.7; runtime 2.4h -> 65.5hImproved
sample base model width from Uniform[64, 512] instead of a fixed width (ablation)TabPack† · 17 tabular datasetsmean rank: 1.6 ± 0.6 -> 1.6 ± 0.5; runtime 2.4h -> 3.5hNo clear effect
use online ensembling instead of offline ensembling (ablation)TabPack · 17 tabular datasetsmean rankImproved
add periodic feature embeddings to the base MLPsTabPack · 17 tabular datasetstask performanceImproved
use Muon instead of AdamW optimizer (ablation)TabPack and baselines · 17 tabular datasetstask performanceImproved
vary number of base models m, width and max depth around default (ablation)TabPack† · 17 tabular datasetsrelative improvement over MLPNo clear effect
replace random hyperparameter diversity with a single tuned config as a source of performance (ablation)TabPack† · 17 tabular datasetsmean rankNo clear effect

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