Hacker News: llmmadnesshttps://news.ycombinator.com/user?id=llmmadnessHacker News RSShttps://hnrss.org/hnrss v2.1.1Sun, 03 May 2026 18:54:27 +0000<![CDATA[New comment by llmmadness in "Even 'uncensored' models can't say what they want"]]>Totally! In some of the cases (we used LLMs to help us generate these) the target word is not clear enough for a human either. So for some of these it turns into more of a guessing game than a flinch measurement.

Agreed, the expectation would be that the flinch measurement becomes stronger. If you are interested in making it better feel free to reach out on the repo!

]]>Tue, 21 Apr 2026 15:25:48 +0000https://news.ycombinator.com/item?id=47850156llmmadnesshttps://news.ycombinator.com/item?id=47850156https://news.ycombinator.com/item?id=47850156<![CDATA[New comment by llmmadness in "Even 'uncensored' models can't say what they want"]]>it's all in the repo. click through to the benchmark it's linked there

]]>Tue, 21 Apr 2026 00:30:27 +0000https://news.ycombinator.com/item?id=47843084llmmadnesshttps://news.ycombinator.com/item?id=47843084https://news.ycombinator.com/item?id=47843084<![CDATA[New comment by llmmadness in "Even 'uncensored' models can't say what they want"]]>We started with a Polymarket project: train a Karoline Leavitt LoRA on an uncensored model, simulate future briefings, trade the word markets, profit. We couldn't get it to work. No amount of fine-tuning let the model actually say what Karoline said on camera. It kept softening the charged word.

]]>Mon, 20 Apr 2026 22:43:29 +0000https://news.ycombinator.com/item?id=47842022llmmadnesshttps://news.ycombinator.com/item?id=47842022https://news.ycombinator.com/item?id=47842022<![CDATA[Even 'uncensored' models can't say what they want]]>Article URL: https://morgin.ai/articles/even-uncensored-models-cant-say-what-they-want.html

Comments URL: https://news.ycombinator.com/item?id=47842021

Points: 178

# Comments: 137

]]>Mon, 20 Apr 2026 22:43:29 +0000https://morgin.ai/articles/even-uncensored-models-cant-say-what-they-want.htmlllmmadnesshttps://news.ycombinator.com/item?id=47842021https://news.ycombinator.com/item?id=47842021<![CDATA[Does an LLM Trained on Epstein's Voice Make Better Deals?]]>Article URL: https://morgin.ai/articles/epsteinbench-we-brought-epsteins-voice-back.html

Comments URL: https://news.ycombinator.com/item?id=47451665

Points: 1

# Comments: 0

]]>Fri, 20 Mar 2026 07:44:41 +0000https://morgin.ai/articles/epsteinbench-we-brought-epsteins-voice-back.htmlllmmadnesshttps://news.ycombinator.com/item?id=47451665https://news.ycombinator.com/item?id=47451665<![CDATA[New comment by llmmadness in "EpsteinBench: We Brought Epstein's Voice Back but Got More Than We Wanted"]]>why not!

]]>Thu, 19 Mar 2026 08:08:45 +0000https://news.ycombinator.com/item?id=47436306llmmadnesshttps://news.ycombinator.com/item?id=47436306https://news.ycombinator.com/item?id=47436306