Early-Signal Detector

method · first seen 5 Oct 2026

mixed-precision quantization

Compressing a neural network by assigning different bit-widths to different modules to balance accuracy and size.

Mentions per week

0619 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

mixture of expertspost-training quantizationKV cache quantizationKV cachetile-based quantizationlooped language modelprefillKullback-Leibler divergenceexpert routing statisticsDNN accelerator mapping