DNeo Diffusion Drafter Speeds Up LLM Inference by Up to 30%

Researchers from NeuralArchitect and Cerebras introduced DNeo, a diffusion-based drafter that speeds up LLM inference by up to 30% via lightweight feature encoding, rejection memory, and architectural improvements to speculative decoding.

2026-11-19 ~ 2026-11-20 · 2 related posts