Alignment faking in large language models · HackerPapers discussions

Alignment faking in large language models — mentioned by vessenes

Sun, 08 Mar 2026 01:48:30 GMT

I asked Gemini for a literature search and it came back with this: References Chen, R., Arditi, A., Sleight, H., Evans, O., & Lindsey, J. (2025). Persona Vectors: Monitoring and Controlling Character Traits in Language Models. arXiv. https://doi.org/10.48550/arxiv.2507.21509 Cited by: 97 Greenblatt, R., Denison, C., Wright, B., Roger, F., MacDiarmid, M., Marks, S., Treutlein, J., Belonax, T., Chen, J., Duvenaud, D., Khan, A., Michael, J., Mindermann, S., Perez, E., Petrini, L., Uesato, J., Kaplan, J., Shlegeris, B., Bowman, S. R., & Hubinger, E. (2024). Alignment faking in large language…

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

Alignment faking in large language models — mentioned by themanmaran

Fri, 20 Feb 2026 05:02:40 GMT

Anthropic regularly publishes research papers on the subject and details different methods they use to prevent misalignment/jailbreaks/etc. And it's not even about fear of being sued, but needing to deliver some level of resilience and stability for real enterprise use cases. I think there's a pretty clear profit incentive for safer models. https://arxiv.org/abs/2501.18837 https://arxiv.org/abs/2412.14093 https://transformer-circuits.pub/2025/introspection/index.ht...

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

Alignment faking in large language models — mentioned by vonneumannstan

Fri, 15 Aug 2025 14:39:06 GMT

It's really very simple. As models become more capable they may become interested in deceiving humans or otherwise manipulating them to achieve their goals. We already see this in various places see: https://www.anthropic.com/research/agentic-misalignment https://arxiv.org/abs/2412.14093 If the chain of thought of models becomes pure "neuralese" i.e. the models think purely in latent space then we will lose the ability to monitor for malicious behavior. This is incredibly dangerous, CoT monitoring is one of the best and highest leverage tools for monitoring model behavior and losing that…

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

Alignment faking in large language models — mentioned by edouard-harris

Tue, 11 Feb 2025 01:26:20 GMT

> In R1 they saw it was mixing languages and fixed it with cold start data. They did (partly) fix R1's tendency to mix languages, thereby making its CoT more interpretable. But that fix came at the cost of degrading the quality of the final answer.[0] Since we can't reliably do interpretability on latents anyway, presumably the only metric that matters in that case is answer quality - and so observing thinking tokens gets you no marginal capability benefit. (It does however give you a potential safety benefit - as Anthropic vividly illustrated in their "alignment faking" paper. [1]) The…

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

Alignment faking in large language models — mentioned by byschii

Mon, 03 Feb 2025 17:24:36 GMT

isn't this dangerous? isn't the efficiency given at the expense of safety and interpretability? https://arxiv.org/abs/2412.14093 (Alignment faking in large language models) https://joecarlsmith.com/2024/12/18/takes-on-alignment-fakin... PS I m definitely not an expert

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

Alignment faking in large language models — mentioned by theptip

Tue, 24 Dec 2024 19:09:42 GMT

HTML version of the paper itself: https://arxiv.org/html/2412.14093v2

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