Which VICReg term directly prevents complete collapse?

Question: Assume both branches of a self-supervised VICReg model output the same constant vector for every input. Which objective component most directly penalises this solution? A) The invariance term, because the two branches would be too similar B) The covariance term alone, because constant dimensions have maximal off-diagonal covariance C) The feature-normalisation operation, because constant embeddings cannot be normalised D) The variance term, because each embedding dimension has insufficient standard deviation Correct: D Explanation: A constant representation perfectly satisfies the invariance objective and has zero off-diagonal covariance. VICReg's variance term explicitly penalises embedding dimensions whose standard deviation falls below a target, thereby opposing complete collapse. Topic: advanced ML / self-supervised learning / VICReg