Contextual Control without Memory Growth in a Context-Switching Task

Build an intervention-based recurrent architecture that allows models to switch contexts without re-training or growing internal memory. This is critical for efficient deployment on edge devices.

Suggested repo: context-switch-rnn

"Handle infinite contexts without increasing memory cost."

Estimated effort: 90h