Unverified50% confidenceFactExact time
A linear probe is an interpretability analysis technique that trains a simple linear classifier on the activations of a certain layer of the model, using classification accuracy to measure whether that layer linearly encodes a certain attribute.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedA linear probe can only detect whether information exists in a linearly separable form within the representation space, and is considered a conservative and reliable standard for measuring whether information is explicitly encoded.80% similarUnverified探测（probing）技术的基本思路是训练一个简单的线性分类器从神经网络某一层的激活值中提取特定信息75% similarUnverifiedLogistic regression probes extract hidden state vectors from intermediate Transformer layers to train linear classifiers, testing whether a given layer encodes specific information.68% similarUnverifiedStatic analysis refers to examining a program without actually running it, in contrast to dynamic analysis, which requires actually executing code to observe behavior.66% similarUnverifiedActivation probes are a class of lightweight monitoring tools, usually a small classifier or transformer, trained to read the internal activation vectors of one or several layers during large model inference, thereby determining whether the model is currently invoking a certain type of specific knowledge or is in a certain intentional state.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1023850API
curl https://kongchang.com/api/v1/knowledge/claims/1023850MCP
get_claim(id=1023850)