Fable 手写 10 万参数模型权重，直接识别真假法律引用
Why we picked it
有人让 Fable 直接手写神经网络权重，绕过训练就做出一个能识别真假法律引用的小模型，路子挺新鲜。
Legora 的 CTO Jacob Lauritzen 让 Fable 手写一个 100k 参数模型的权重，用于区分真实与虚假的法律引用。这次尝试相当于一次成型的完整神经网络，效果优于用训练方式得到的模型。文中将这类做法称为 weight-space programming（权重空间编程）。
Original · elvis
🤯Another mind-blowing example (this time from the CTO of Legora) on Fable hand-writing weights of a model. This weight-space programming stuff is seriously some of the most exciting research I have seen in a long time. Jacob Lauritzen @jacsebl Got fable to hand-write weights of a 100k param model to recognize real versus fake legal citations. Basically one-shotted a full NN that’s better than training it 🔗 View Quoted Tweet 💬 4 🔄 1 ❤️ 11 👀 1993 📊 5 ⚡
