Guidelabs launches Causal Diffusion Engine, adding interpretability to diffusion LLMs with no latency cost

juliusadml · x · 2026-10-30

Guidelabs introduced the Causal Diffusion Engine (CDE), the inference engine powering the interpretable diffusion LLM Steerling-8B. CDE attaches concept explanations and steering to every generated token without added latency, arguing interpretability doesn't have to slow down inference.

Original post →

Guidelabs launches Causal Diffusion Engine, adding interpretability to diffusion LLMs with no latency cost

More from Models