Hacker News new | ask | show | jobs
by mostin 505 days ago
I think the ablated models are really interesting as well: https://huggingface.co/bartowski/deepseek-r1-qwen-2.5-32B-ab...

For some reason I always get the standard rejection response to controversial (for China) questions, but then if I push back it starts its internal monologue and gives an answer.