Attractors in end-to-end neural diarization may not require encoding speaker characteristic information.