Consistency diffusion language models: Up to 14x faster, no quality loss from Hacker News on 2026-02-20 04:15 (#73PV3) Comments
