Views
No views yet
W_new = W × (1 + α × mask)mask identifies weights important for creative tasks but not for routine/associative tasks.1from transformers import AutoModelForCausalLM, AutoTokenizer
2
3model = AutoModelForCausalLM.from_pretrained("priorcomputers/phi-3-small-8k-instruct-cn-dat-kr0.1-a0.5-creative", trust_remote_code=True)
4tokenizer = AutoTokenizer.from_pretrained("priorcomputers/phi-3-small-8k-instruct-cn-dat-kr0.1-a0.5-creative", trust_remote_code=True)
5
6# Use like any other model
7outputs = model.generate(...)1@misc{creativityneuro2025,
2 title={CreativityNeuro: Mechanistic Interpretability for LLM Creativity},
3 author={Prior Computers},
4 year={2025},
5 url={https://huggingface.co/priorcomputers}
6}