Yann LeCun 谈 JEPA 世界模型：用理解替代知识堆叠，降低参数与内存需求
LLMs accumulate a lot of declarative knowledge. That's how they can answer a large number of questio...
LeCun 本人下场解释 JEPA 和 LLM 的根本区别，还附了论文，想搞懂世界模型路线的看这条就够。
Yann LeCun 在 X 上发文指出，LLM 通过参数堆积陈述性知识来回答海量问题，因此参数量和内存需求巨大。JEPA 世界模型改用理解来替代知识堆叠，所需内存和训练数据明显更少。他转发的研究论文显示，在 SIGReg 或 VCReg 等抗坍塌判据（而非重建式训练）下，文本或图像数据的底层相似性结构会唯一决定表示点云的形状。
LLMs accumulate a lot of declarative knowledge. That's how they can answer a large number of questio...
LLMs accumulate a lot of declarative knowledge. That's how they can answer a large number of questions. That's also why they require so many parameters and hence so much memory. But one can replace the accumulation of knowledge by *understanding* That what JEPA world models do. That's one reason they can be useful with considerably less memory and training data than LLMs. This paper merely shows that the underlying similarity structure of the data, whether text or image (e.g. co-occurence of words in text or objects in images) uniquely determines the shape of the representation point cloud, when the encoder is trained trained with a common anti-collapse criterion like SIGReg or VCReg (not by reconstruction). 💬 143 🔄 324 ❤️ 3225 👀 254038 📊 610 ⚡
