NLPCC 2026 Task Focuses on Difficulty-Aware Medical Video Understanding

The NLPCC 2026 conference is launching a new shared task called Difficulty-Aware Medical Instructional Video Question Answering (DA-MIVQA). This initiative builds upon previous challenges by introducing a more nuanced evaluation of AI systems' ability to understand medical instructional videos across multiple languages and modalities. A key innovation of DA-MIVQA is its explicit distinction between simple and complex questions. Simple questions can often be answered using text from subtitles, while complex ones demand visual grounding, comprehension of procedures, and the integration of information from both visual and textual sources. This approach aims to provide a more comprehensive assessment of AI's reasoning capabilities. The challenge includes three tracks: Difficulty-Aware Temporal Answer Grounding in Single Video, Difficulty-Aware Video Corpus Retrieval, and Difficulty-Aware Temporal Answer Grounding in Video Corpus. The dataset, sourced from public medical instructional channels, covers diverse scenarios like first aid and nursing, with manual difficulty annotations to ensure robust evaluation.