Unverified50% confidenceFactExact time
Contrastive activation probing uses cosine similarity to measure the directional deviation of intermediate-layer activations between the base model and the fine-tuned model under the same probe prompts, with lower similarity indicating greater drift.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedCosine similarity and Euclidean distance may give diametrically opposite judgments about which two vectors are more similar.73% similarUnverifiedCosine similarity measures how close two vectors are in direction, with a value range of -1 to 1; the closer the value is to 1, the more similar the semantics, and because it is insensitive to vector magnitude and stable to compute, it has become the most mainstream choice in RAG scenarios.72% similarUnverifiedLater checkpoints may have stronger similarity but can also overfit and respond poorly to prompts; earlier checkpoints are more flexible but may not have fully learned the identity features.70% similarUnverifiedThe study theoretically derives that when the encoder merges the distinguishable outcomes of different actions under the same state as well as different appearances, a positive lower bound exists on the squared inverse prediction error.69% similarUnverifiedCosine similarity measures the angle between two vectors, where a result of 1 means the same direction (the same meaning) and 0 means irrelevant; it ignores vector length, so a short question can also match a long document.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1070464API
curl https://kongchang.com/api/v1/knowledge/claims/1070464MCP
get_claim(id=1070464)