The Video Browser Showdown (VBS) provides a rigorous benchmark for evaluating interactive video retrieval systems under strict real-time constraints. While traditional tasks such as Known-Item Search (KIS) and Ad-hoc Video Search (AVS) remain central, the recent introduction of Visual Question Answering (VQA) has introduced new challenges. Participants are required to identify fleeting answer-bearing moments within long video sequences, often under ambiguous queries and limited time. Current systems face three main limitations: difficulties in accurately localizing these brief segments, the high verification effort demanded from users under time pressure, and limited capability to effectively integrate diverse modalities such as visual content and speech. To address these challenges, we present the NII-UIT system for VBS 2026, explicitly designed to advance VQA in interactive video retrieval. Our framework introduces an Answer Span Prediction module to highlight candidate temporal regions, a Candidate Answer Suggestion mechanism that aggregates multimodal cues to generate verifiable answer options, and a dedicated In-Video Retrieval component for frame-level evidence discovery. Complementary system optimizations and an enhanced user interface further improve interaction speed and reduce cognitive load. By shifting the user’s role from open-ended searching to guided verification, our contributions are expected to improve efficiency and accuracy, offering a practical step toward more effective VQA systems in competitive evaluation scenarios.