CycleGAN, a Master of Steganography · HackerPapers discussions

CycleGAN, a Master of Steganography — mentioned by alwa

Wed, 07 Oct 2026 15:49:50 GMT

https://arxiv.org/abs/1712.02950 (2017) …if you’re curious and missed that one like I did. Snack-sized paper with lots of satisfying visual examples.

Paper and discussion history. A mention is not an endorsement.

CycleGAN, a Master of Steganography — mentioned by nneonneo

Thu, 05 Mar 2026 10:12:53 GMT

Somewhat annoyingly, there's been research that suggests that models can pass information to each other via (effectively) steganographic techniques - specific but apparently harmless choices of tokens, wordings, and so on; see https://arxiv.org/abs/1712.02950 and https://alignment.anthropic.com/2025/subliminal-learning/ for some simple examples. While it feels unlikely that a simple "write this spec from this code" + "write this code from this spec" loop would actually trigger this kind of hiding behaviour, an LLM trained to accurately reproduce code from such a loop definitely would be…

Paper and discussion history. A mention is not an endorsement.

CycleGAN, a Master of Steganography — mentioned by Calavar

Fri, 08 Aug 2025 11:24:47 GMT

> Researchers had observed similar patterns in BERT, where "a surprisingly large amount of attention focuses on the delimiter token [SEP] and periods," which they argued was used by the model as a sort of no-op. The same summer at Meta, researchers studying vision transformers found similar behavior, observing that models would repurpose uninformative background patches as computational scratchpads. This seems to go beyond just transformers. For example, I recall reading a paper a while ago that showed a similar effect in an image to image model with a GAN/U-Net architecture [1]. [1]…

Paper and discussion history. A mention is not an endorsement.

CycleGAN, a Master of Steganography — mentioned by wizzwizz4

Sat, 07 Sep 2024 21:14:10 GMT

> You could even get GPT to generate it This isn't a good idea, if you want a fair test. See https://gwern.net/doc/reinforcement-learning/safe/2023-krako..., specifically https://arxiv.org/abs/1712.02950.

Paper and discussion history. A mention is not an endorsement.

CycleGAN, a Master of Steganography — HN submission

Thu, 03 Jan 2019 21:48:23 GMT

CycleGAN, a Master of Steganography

3 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

CycleGAN, a Master of Steganography — HN submission

Wed, 02 Jan 2019 10:15:24 GMT

CycleGAN, a Master of Steganography [2017]

3 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.