# Preprint reports faster GB200 training with polynomial function replacements

_Published Saturday, October 3, 2026 at 12:14 AM EDT · AI, Science · Latest · Tier 2 — Notable_

Researchers report in an arXiv preprint that replacing selected mathematical functions with short polynomial calculations improved complete training-step throughput on GB200 hardware by 2.7%, 2.9%, and 8.0% across three tasks.

The replacements approximate functions used inside language models, combining symmetry, rounding to the target number format, and packed arithmetic within the kernels that consume their outputs. A fourth task, sigmoid attention, improved complete-attention forward performance by 7.4% but the complete GPU step by 0.3%.

The evaluation included one paired pretraining comparison per task. At common horizons near 100 billion tokens, final smoothed training-loss differences between polynomial and native implementations ranged from -0.107 to +0.079, showing that the substitutions also changed model training behavior.

## Sources

- [cs.LG updates on arXiv.org](https://arxiv.org/abs/2610.00049)

---
Canonical: https://techandbusiness.org/newswire/lhe6dxtPBXsJZft0rX1qaP
Published: 2026-10-03T04:14:11.013Z
Story chronology: 2026-10-03T04:00:00.000Z
Retrieved: 2026-10-03T06:26:21.658Z
Publisher: Tech & Business (techandbusiness.org)