Unverified50% confidenceFactExact time
The theoretical foundation of abliteration comes from the field of representation engineering, which has found that certain high-level semantic concepts within a model can be identified as specific linear directions in activation space
1
Sources
50%
Confidence
Long-term
Relevance
9/18/2026
First Seen
Sources
Related Entities
Related Claims
Unverified表征工程核心洞见是大语言模型的高层语义概念在内部激活空间中呈现近似线性的几何结构，即存在概念方向向量76% similarUnverifiedThe core principle of Abliteration is to identify the refusal direction in the model's residual stream and eliminate it via orthogonal projection from the weight matrices.73% similarUnverifiedConcept Bottleneck Models introduce a concept layer between the input layer and output layer, mapping the high-dimensional feature space to a human-understandable concept space71% similarUnverified线性表征假说认为高级语义概念以线性方向编码在残差流中，是当前可解释性研究的核心假设之一69% similarUnverified分词特征差异发生在推理链路底层，难以通过提示工程掩盖，可作为模型识别手段68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/933329API
curl https://kongchang.com/api/v1/knowledge/claims/933329MCP
get_claim(id=933329)