https://hackerpaper.kaipfister.de/p/2512.17075/Collected HN comments and submissions linking this paper.enWed, 04 Nov 2026 05:35:47 GMThttps://news.ycombinator.com/item?id=50343015urn:hackerpapers:mention:2512.17075:50343015Mon, 02 Nov 2026 20:14:10 GMT<p>In a self-supervised setting you don&#x27;t need ground truth data, as long as you have a proxy task that you can train on which forces the model to learn something useful. Like cutting out bits of images and telling the model to reconstruct the gaps (i.e. MAEs). There are also techniques to watermark model weights to confirm what training data were used: https://arxiv.org/html/2512.17075 I&#x27;m not an expert in this domain, but it seems like there are lots of pieces that could be used to (a) train a model to infer whether some data were trained on and (b) train a model to understand which weights…</p><p><a href="https://hackerpaper.kaipfister.de/p/2512.17075/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.CLhttps://news.ycombinator.com/item?id=46693617urn:hackerpapers:submission:2512.17075:46693617Tue, 20 Jan 2026 16:19:19 GMT<p>Perturb Your Data: Paraphrase-Guided Training Data Watermarking</p><p></p><p>1 point; 0 discussion replies on HN. <a href="https://hackerpaper.kaipfister.de/p/2512.17075/">Paper and source history</a>. A submission is not an endorsement.</p>cs.CL