T2Quant optimizes diffusion language model quantization via temporal awareness
October 23, 2026
T2Quant is a mixed-precision quantization framework designed for Diffusion Language Models that allocates activation precision based on task and denoising step sensitivity. It achieves accuracy gains of up to 6.7 percentage points across three DLMs by predicting precision requirements without ground-truth labels.
HOW THIS AFFECTS YOU
●
builderYou can improve the efficiency of deploying diffusion-based text models on constrained hardware.
●
researcherThe framework demonstrates that quantization sensitivity in DLMs is a predictable function of both task and denoising time.