Diffusion‑based LLMs that generate many parallel tokens rather than one‑by‑one

(inceptionlabs.ai) by binyu | Jun 20, 2026 | 0 comments on HN