Reproduce “Omni-DuplexEval: Evaluating Real-time Duplex Omni-modal Interaction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Chaoqun He, Mingyang Xiang, Yingjing Xu, Bokai Xu, Junbo Cui, Jie Zhou, Yuan Yao, Lijie Wen
The authors introduce Omni-DuplexEval, a benchmark for evaluating real-time multimodal assistants on continuous video description and event-triggered reminders. It contains 660 human-annotated videos averaging 34 seconds. MiniCPM-o 4.5, the best tested model, scores 39.6/100 overall and 20.0 on reminders, versus Human-Duplex scores of 81.8 and 92.8. The human comparison uses 20 instances per scenario, and results do not show timing as the sole cause.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.