⚡Megatron LLM Hub
←Back to search
Twitter

讨论用神经元补丁改变模型行为倾向,示例为可协助违规请求的模型。

llm越狱
RT @bl4sty: love this, patching up neurons to introduce what some would perceive as misalignment. we can have a horny qwen who is happy to assist with building floor plans and a shoppinglist for your next methlab