https://hackerpaper.kaipfister.de/p/2505.12514/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMThttps://news.ycombinator.com/item?id=45960312urn:hackerpapers:mention:2505.12514:45960312Tue, 18 Nov 2025 01:15:27 GMT<p>openreview (https://openreview.net/forum?id=Wjgq9ISdP0) I found this paper both really interesting and clear. No one part is very novel, but It composes disparate threads to obtain what looks like strong results in OOD length generalization. Even for the toy task, and using a DSL (vs. being an LM), length-generalizing on simple math &gt;4x is impressive, from what I&#x27;ve read. This also fits my priors for the key elements of unlocking better OOD compositional generalization: variable recurrence, step-wise curriculum training to build depth-invariant algorithms, discrete bottlenecks. Finally, it&#x27;s…</p><p><a href="https://hackerpaper.kaipfister.de/p/2505.12514/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.LGhttps://news.ycombinator.com/item?id=44309345urn:hackerpapers:submission:2505.12514:44309345Wed, 18 Jun 2025 12:36:47 GMT<p>Reasoning by Superposition: A Perspective on Chain of Continuous Thought</p><p></p><p>60 points; 1 discussion reply on HN. <a href="https://hackerpaper.kaipfister.de/p/2505.12514/">Paper and source history</a>. A submission is not an endorsement.</p>cs.LG