AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

End-to-End Compression for Tabular Foundation Models

Guri Zabërgja, Rafiq Kamel, Arlind Kadra, Christian Frey, Josif Grabocka

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Language model pretraining and induction · 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
compress training context with learned TACO compressor (1% rate) instead of full contextPOT predictor transformer · Synthetic 15k x 500, no KV cachesubsequent predict time: 28.67s -> 306msImproved
compress context to 1% with TACOPOT predictor transformer · Synthetic 15k x 500, no KV cachepredict memory: 22.45GB -> 549MB (-97.6%)Improved
compress context to 1% with TACOPOT predictor transformer · TabArena classification (26 datasets)mean ROC-AUC: 0.862 -> 0.855No clear effect
add TACO compression to enable KV cachingPOT with KV cache · Synthetic grid N 5k-50k, M 100-1000feasibility (OOM): POT OOM on 32 of 100 grid datasetsImproved
TACO compression raises fit-time memory (no KV cache)TACO · Synthetic 15k x 500fit memory: 36MB -> 92MBHurt
train compressor and predictor jointly instead of freezing predictor at TabPFN v2.0 weights (ablation)TACO · TabArena classificationROC-AUCImproved
train one model with compression rate sampled from {1,2,4,8,16}% instead of rate-specific models (ablation)TACO · TabArena classificationROC-AUCNo clear effect
use learned TACO compressor instead of random or kNN subsampling of context (ablation)POT · TabArena classification, 1%-16% ratesROC-AUCImproved
chunk-and-stitch compression with TACO instead of random/kNN subsamplingPOT / TabPFNv2 · MetroPT-3AUPRC: 0.7293 -> 0.8955 (vs POT Random)Improved
higher compression rate (more retained rows) (ablation)TACO vs baselines · TabArena classificationROC-AUC gapMixed
higher compression rate 1% -> 16% (ablation)TACO · Synthetic 15k x 500, no KV cachesubsequent predict speedup: ×93.6 -> ×15Hurt

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