https://hackerpaper.kaipfister.de/p/2409.14586/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMThttps://news.ycombinator.com/item?id=41739850urn:hackerpapers:mention:2409.14586:41739850Fri, 04 Oct 2024 10:12:28 GMT<p>I think this might be key, in addition to some landmark tokens to quickly backtrack to. The big question is how to train such model. There is a recent paper from Meta that propose a way to train a model to backtrack its generation to improve generation alignment [0]. [0] https://arxiv.org/html/2409.14586v1</p><p><a href="https://hackerpaper.kaipfister.de/p/2409.14586/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.LG