@misc{indiciae106274d83843, title = {Disentangling Channel Semantics in Vision Transformers via Token Decorrelation and Composition-Aware Modulation}, author = {Daeun Kim and Hyesong Choi and Hyejin Park and Dongbo Min}, year = {2026}, url = {https://searcharxiv.com/abs/2611.10497}, note = {Source identifier: 2611.10497} }