Does YOLO give preference to color over shape or vice-versa while detecting an object?

YOLO has 3 input channels (RGB typically) and is a CNN, a neural net that uses stacked (trainable) convolution filters. Single convolution filters act as feature detectors and can be (trained to be) sensitive for straight edges from dark to light or a gradual passage from green to red. By stacking those you can build filters for more complicated features (such as faces).

TL;DR The network can learn to differentiate green and red circles if you want to.