https://hackerpaper.kaipfister.de/p/2602.06036/Collected HN comments and submissions linking this paper.enTue, 10 Nov 2026 07:51:04 GMThttps://news.ycombinator.com/item?id=50428960urn:hackerpapers:mention:2602.06036:50428960Mon, 09 Nov 2026 17:54:58 GMT<p>https://arxiv.org/abs/2602.06036 - they use a faster (single-pass) diffusion model to generate predictions for 7 to 15 tokens in a single step. Then if they&#x27;re accepted that really speeds up the autoregressive model because it doesn&#x27;t have to generate tokens one at a time.</p><p><a href="https://hackerpaper.kaipfister.de/p/2602.06036/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.CL