English

QuIDE: Mastering the Quantized Intelligence Trade-off via Active Optimization

Machine Learning 2026-05-13 v1 Artificial Intelligence

Abstract

There is currently no unified metric for evaluating the efficiency of quantized neural networks. We propose QuIDE, built around the Intelligence Index I = (C x P)/log_2(T+1), which collapses the compression-accuracy-latency trade-off into a single score. Experiments across six settings -- SimpleCNN (MNIST, CIFAR), ResNet-18 (ImageNet-1K), and Llama-3-8B -- show a task-dependent Pareto Knee. 4-bit quantization is optimal for MNIST and large LLMs, while 8-bit is the sweet spot for complex CNN tasks (ResNet-18 on ImageNet), where 4-bit PTQ collapses accuracy catastrophically. The accuracy-gated variant I' correctly flags these non-viable configurations that the raw I would reward. QuIDE provides a reproducible evaluation protocol and a ready-to-use fitness function for mixed-precision search.

Keywords

Cite

@article{arxiv.2605.10959,
  title  = {QuIDE: Mastering the Quantized Intelligence Trade-off via Active Optimization},
  author = {Xiantao Jiang},
  journal= {arXiv preprint arXiv:2605.10959},
  year   = {2026}
}

Comments

16 pages, 9 figures