Verified75% confidenceFactExact time
Activation steering is an inference-time intervention technique that requires no model retraining: at each token generation step, a scaled direction vector is added to the hidden state of a specific layer (h ← h + α·v), where α controls the intervention strength and v is the target direction.
3
Sources
75%
Confidence
Long-term
Relevance
10/23/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedActivation steering is a technique that intervenes in the model's internal hidden-layer representations at inference time, with representative methods including contrastive activation addition (CAA) and inference-time intervention (ITI).77% similarUnverifiedActivation steering or representation engineering applies intervention vectors directly to the model's internal activation layers, altering the generation direction without modifying the model's weights.76% similarUnverifiedDuring the pretraining phase, the model's objective function is typically to maximize the log-likelihood of the next token and does not include signals for refusing to answer.72% similarUnverifiedAlignment training creates a 'refusal tendency' inside the model; this tendency is not implemented through external prompts but is encoded into the model weights, which is why it is called an 'internal constraint.'72% similarVerifiedAdversarial distillation attackers systematically send large volumes of requests to the target model, collecting input-output pairs—especially responses containing complete reasoning chains—to train their own models and replicate the target model's capabilities.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1088144API
curl https://kongchang.com/api/v1/knowledge/claims/1088144MCP
get_claim(id=1088144)