Perturbation: A simple and efficient adversarial tracer for representation learning in language models · HackerPapers discussionshttps://hackerpaper.kaipfister.de/p/2603.23821/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMTPerturbation: A simple and efficient adversarial tracer for representation learning in language models — mentioned by jsroznerhttps://news.ycombinator.com/item?id=49533463urn:hackerpapers:mention:2603.23821:49533463Wed, 02 Sep 2026 08:33:05 GMT<p>A big problem with some of these supervised* interpretability approaches is that they can find spurious structure. (There are lots of ways to make the model do what you want; which is roughly what Hewitt and Liang 2019 showed). This paper draws a contrast to a previous method, DAS (distributed alignment search) on page 20. These and related methods rest on theories of causal abstraction, which are great in theory, but harder in practice. DAS, for example, has faced numerous recent criticisms (Makelov 2024, Meloux 2025, Sutter 2025, Grant 2026, Kumon 2026). My favorite is the quite…</p><p><a href="https://hackerpaper.kaipfister.de/p/2603.23821/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.CL