←back to thread

Heretic: Automatic censorship removal for language models

(github.com)
745 points melded | 1 comments | 16 Nov 25 15:00 UTC | HN request time: 0.21s | source
Show context
richstokes ◴[16 Nov 25 17:52 UTC] No.45946953[source]▶
Is there a way to use this on models downloaded locally with ollama?
replies(2): >>45947557 #>>45949300 #
1. EagnaIonat ◴[16 Nov 25 19:12 UTC] No.45947557[source]▶
A lot of the models in Ollama you can already easily bypass safe guards without having to retrain. OpenAI's open source models can be bypassed just by disabling thinking.