Turning Your Strength into Watermark: Watermarking Large Language Model via Knowledge Injection · HackerPapers discussionshttps://hackerpaper.kaipfister.de/p/2311.09535/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMTTurning Your Strength into Watermark: Watermarking Large Language Model via Knowledge Injection — mentioned by jruohonenhttps://news.ycombinator.com/item?id=38299506urn:hackerpapers:mention:2311.09535:38299506Fri, 17 Nov 2023 04:17:30 GMT<p>I remain skeptical because it is easy enough to pass the output from a LLM through a filter, a scrubber, another LLM, or whatever. Furthermore, people (humans) typically alter the outputs as well. Here is another fresh paper on the same topic from Chinese researchers: https://arxiv.org/abs/2311.09535 While they discuss some attacks, they too do not consider altering the output from a LLM, possibly through f(g(h(...))).</p><p><a href="https://hackerpaper.kaipfister.de/p/2311.09535/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.CR