Niklast

joined 2 years ago
[–] Niklast@lemmy.world 22 points 5 months ago

They train grok like other llm Models, but they use system prompts to manipulate answers. That way they can tweak it quickly without retraining.