Training-Free Decoder Boosts Diffusion LLM Throughput Up to 14x
STDec is a training-free decoder that exploits spatial and temporal stability in diffusion language models to deliver up to 14x throughput gains without degrading accuracy or modifying model weights.
Read the full story at AlphaSignal ↗
Timeline · 1 report
- 2026-11-12 06:01 · AlphaSignal
STDec Makes Diffusion LLMs 14x Faster Without Touching Model Weights