SVFSearch: A Multimodal Knowledge-Intensive Benchmark for Short-Video Frame Search in the Gaming Vertical Domain

In three linesSVFSearch is a multimodal benchmark for short-video frame search in the Chinese gaming domain. It contains 5,000 test examples and 4,198 training examples based on real game scenes. Evaluation compares direct QA, RAG, Plan-Act-Replan agents, and learned search models: best open-source model reaches 66.4%, best practical agent 79.1%, oracle 95.4%.

Summary generated by Claude — human-verified