Unverified50% confidenceFactExact time
Large language models encode semantic concepts as specific directional vectors in a high-dimensional representation space, with refusal behavior corresponding to an identifiable direction.
1
Sources
50%
Confidence
Long-term
Relevance
9/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMultimodal large language model agent methods have strong planning capabilities and can directly drive scene construction using natural language, but lack precise 3D spatial control.74% similarUnverified大语言模型在符号操作、模式识别、定义比对等任务上表现较强，但在需要深层数学直觉和构造性证明的创造性工作上仍相当薄弱74% similarUnverifiedLarge language models are essentially stateless functions, and a model's memory depends entirely on the content within the context window.73% similarUnverifiedMainstream text embedding models are typically trained based on contrastive learning, mapping semantically similar text pairs to nearby positions in vector space and pushing unrelated text pairs further apart.72% similarUnverified研究者在分析多语言BERT（mBERT）内部表示时发现，即便没有平行语料的明确监督，模型也能自发学到不同语言相同概念在向量空间中高度相似的对齐结构70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/861882API
curl https://kongchang.com/api/v1/knowledge/claims/861882MCP
get_claim(id=861882)