Controlled Dynamics Attractor Transformer

In three linesCDAT couples Transformer attention with continuous attractor neural network (CANN) dynamics. The model combines von Mises-Fisher attention energy with Hopfield refinement and excitation-inhibition modulation. Achieves state-of-the-art results on graph anomaly detection and graph classification benchmarks.

Summary generated by Claude — human-verified