VaQuitA: Enhancing Alignment in LLM-Assisted Video Understanding