llm_steer

So if I understand, this is basically advanced activation steering as a service? And you have already identified vectors for several open models that make them more truthful or better at reasoning and apply them automatically? Because the API has a persona option which might be achieved with something like this https://github.com/Mihaiii/llm_steer or maybe for closed models you just have to append to the prompt. What open source models are available? In the docs I only see mention of Google Flash Lite or something which is closed.

What's opaque about mistral large? https://huggingface.co/mistralai/Mistral-Large-Instruct-2407 I have total control over the model, literally dozens of generation parmaters to play with, literally thousands of open source projects like this https://github.com/Mihaiii/llm_steer or https://colab.research.google.com/drive/1a-aQvKC9avdZpdyBn4j... which allow me to jailbreak the model to my hearts content - except I didn't need to do that since Mistral isn't actually aligned anyway! While you're the one relying on "opaque" knowledge agents, I will have DPO'd my open access model with my custom huge debate evidence dataset to be a better competitive debater than any living human on the planet. Good luck.

Hey, guys! I'm pretty amazed steering vectors work so well so I made a Python module to add steering vectors more easily. Google Colab link included in the repo. Let me know what you think.