Unverified50% confidenceFactExact time
SAGA (Structure-Aware Guided Alignment) is a training-only regularization method that aligns neural molecular representations with fingerprint-based Tanimoto similarity via centered kernel alignment (CKA).
1
Sources
50%
Confidence
Long-term
Relevance
11/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe researchers propose ScaleBrain, a scale-consistent pretraining framework that explicitly coordinates multi-scale representations from both the model input and the optimization objective.65% similarUnverifiedNeural representations trained by overfitting to a single shape are called overfitting-based or per-shape fitting methods, while methods with generalization capability are called feed-forward methods (e.g., DeepSDF, OccNet).62% similarUnverifiedSparse autoencoders (SAEs) are unsupervised learning tools trained to decompose a neural network's activation vectors into a set of sparse, interpretable feature directions.60% similarUnverified训练数据归因(TDA)的常见技术路径包括影响函数、相似度检索和成员推断攻击60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1179361API
curl https://kongchang.com/api/v1/knowledge/claims/1179361MCP
get_claim(id=1179361)