Unverified50% confidenceFactExact time
TCAV (Testing with Concept Activation Vectors) quantifies the influence of a given concept on prediction results by training a linear classifier in the activation space of the model's intermediate layers.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTest-Time Adaptation (TTA) allows a model to perform online updates of batch normalization layer parameters using a small amount of target domain data during inference, serving as a lightweight approach to addressing domain shift.70% similarUnverifiedThis study published on arXiv proposes a new proposition called Alignment Forecasting: predicting before training begins whether a given fine-tuning dataset will significantly exacerbate specific misalignment behaviors.68% similarUnverifiedA linear probe is a simple linear classifier trained on model activations, used to detect whether specific behaviors or concepts exist inside the model.68% similarUnverifiedSelecting the already-trained model whose parameters are closest to a specific trade-off point to approximate that point is a method with some effectiveness.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1001895API
curl https://kongchang.com/api/v1/knowledge/claims/1001895MCP
get_claim(id=1001895)