Unverified50% confidenceOpinionExact time
Transferability is a key metric for evaluating such neuron modulation methods: neuron directions in different models (even different sizes or fine-tuned versions of the same architecture) are not necessarily aligned, and a persona direction found in one model may fail or produce opposite effects when transferred to another model.
1
Sources
50%
Confidence
Long-term
Relevance
10/28/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe researchers propose that what actually transfers from written instructions to actual operation is not the sequence of steps but a mental model of the system's internal mechanism, that is, conditional knowledge of the form 'if X is in a certain state, then what will happen to Y'.70% similarUnverifiedThe transferability of adversarial attacks refers to the ability of adversarial examples generated for one model to mislead another model with a different structure or different training data with high probability, arising from similarities in the decision boundaries of different models.69% similarUnverifiedTo verify the transferability of the theory to neural network models, the researchers use a counterfactual proxy to estimate the influence of spurious-correlation dependence and use the generalized Gauss-Newton approximation to estimate the curvature of the language model.69% similarUnverifiedModel routing is essentially a scheduling middleware layer that transfers the cognitive burden of model selection from humans to the system.69% similarUnverified预训练模型迁移复用的假设是模型在源域学到的特征具有跨域泛化能力，但高层语义特征可能存在显著的域偏移69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1108621API
curl https://kongchang.com/api/v1/knowledge/claims/1108621MCP
get_claim(id=1108621)