CourseTimeQA: A Lecture-Video Benchmark and a Latency-Constrained Cross-Modal Fusion Method for Timestamped QA
Researchers have developed a new benchmark, CourseTimeQA, for timestamped question answering over educational lecture videos. They also introduced a lightweight, latency-constrained cross-modal retrieval method called CrossFusion-RAG. This method combines frozen encoders with a learned projection and cross-attention mechanisms to achieve improved performance and low latency on a single GPU. AI
IMPACT Introduces a new benchmark and method for timestamped QA, potentially improving educational AI tools.