description [NeurIPS 2025][LLM Pretraining][generalization] This paper proposes Gradient-Weight Alignment (GWA), which quantifies the directional consistency (cosine similarity) between the gradient ...
The modulatory effect of evaluative valence on fear generalization in social anxiety: an SSVEP study
Social anxiety is characterized by excessive sensitivity and concern about social evaluation. While previous research has demonstrated attentional bias and fear generalization in socially anxious ...
If you want to improve your aerobic capacity, play full-court basketball, not softball. To improve your analytical skills, learn to play chess or bridge, not Chutes and Ladders. If you really want to ...
Conceptual overview of this work. (A) Animals have goals, which they must learn to achieve. In this case, consider revising a paper. (B) Some feedback signals (e.g., from reviewers or co-authors) will ...
Gradient descent-trained neural networks operate effectively even in overparameterized settings with random weight initialization, often finding global optimum solutions despite the non-convex nature ...
Grokking is a newly developed phenomenon where a model starts to generalize well long after it has overfitted to the training data. It was first seen in a two-layer Transformer trained on a simple ...
Abstract: The rapid development of face forgery technology has posed a significant threat to information security. While deepfake detection has proven to be an effective countermeasure, it often ...
Abstract: Deep neural networks (DNN) have demonstrated unprecedented success for various applications. However, due to the issue of limited dataset availability and the strict legal and ethical ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results