Beyond Monocular Vision: Assessing LLaVA's Performance on an Augmented CLEVR-like Dataset with Binocular Images