The Devil Is in Gradient Entanglement: Energy-Aware Gradient Coordinator for Robust Generalized Category Discovery

Build an 'Energy-Aware Gradient Coordinator' library that prevents gradient interference in multi-task settings. This allows researchers to train robust models that categorize both known and unknown classes without optimization collapse.

Suggested repo: grad-sync

"Stop gradient entanglement from sabotaging your multi-task models."

Estimated effort: 50h