Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Gangwani, Ishaan, Bansal, Aayam
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2512.00888
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912766788370432
author Gangwani, Ishaan
Bansal, Aayam
author_facet Gangwani, Ishaan
Bansal, Aayam
contents Zero-shot foundation models (FMs) promise training-free prediction on tabular data, yet their hardware footprint remains poorly characterized. We present a fully reproducible benchmark that reports test accuracy together with wall-clock latency, peak CPU RAM, and peak GPU VRAM on four public datasets: Adult-Income, Higgs-100k, Wine-Quality, and California-Housing. Two open FMs (TabPFN-1.0 and TabICL-base) are compared against tuned XGBoost, LightGBM, and Random Forest baselines on a single NVIDIA T4 GPU. The tree ensembles equal or surpass FM accuracy on three datasets while completing full-test batches in <= 0.40 s and <= 150 MB RAM, using zero VRAM. TabICL achieves a 0.8 percentage-point gain on Higgs but requires roughly 40,000 times more latency (960 s) and 9 GB VRAM. TabPFN matches tree-model accuracy on Wine and Housing but peaks at 4 GB VRAM and cannot process the full 100k-row Higgs table. These results quantify the substantial hardware-versus-accuracy trade-offs in current tabular FMs and provide an open baseline for future efficiency-oriented research.
format Preprint
id arxiv_https___arxiv_org_abs_2512_00888
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Light-Weight Benchmarks Reveal the Hidden Hardware Cost of Zero-Shot Tabular Foundation Models
Gangwani, Ishaan
Bansal, Aayam
Machine Learning
Artificial Intelligence
Zero-shot foundation models (FMs) promise training-free prediction on tabular data, yet their hardware footprint remains poorly characterized. We present a fully reproducible benchmark that reports test accuracy together with wall-clock latency, peak CPU RAM, and peak GPU VRAM on four public datasets: Adult-Income, Higgs-100k, Wine-Quality, and California-Housing. Two open FMs (TabPFN-1.0 and TabICL-base) are compared against tuned XGBoost, LightGBM, and Random Forest baselines on a single NVIDIA T4 GPU. The tree ensembles equal or surpass FM accuracy on three datasets while completing full-test batches in <= 0.40 s and <= 150 MB RAM, using zero VRAM. TabICL achieves a 0.8 percentage-point gain on Higgs but requires roughly 40,000 times more latency (960 s) and 9 GB VRAM. TabPFN matches tree-model accuracy on Wine and Housing but peaks at 4 GB VRAM and cannot process the full 100k-row Higgs table. These results quantify the substantial hardware-versus-accuracy trade-offs in current tabular FMs and provide an open baseline for future efficiency-oriented research.
title Light-Weight Benchmarks Reveal the Hidden Hardware Cost of Zero-Shot Tabular Foundation Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2512.00888