{"ID":2834896,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2512.00888","arxiv_id":"2512.00888","title":"Light-Weight Benchmarks Reveal the Hidden Hardware Cost of Zero-Shot Tabular Foundation Models","abstract":"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 \u003c= 0.40 s and \u003c= 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.","short_abstract":"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, Hig...","url_abs":"https://arxiv.org/abs/2512.00888","url_pdf":"https://arxiv.org/pdf/2512.00888v2","authors":"[\"Ishaan Gangwani\",\"Aayam Bansal\"]","published":"2025-11-30T13:17:08Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.AI\"]","methods":"[]","has_code":false}