Reproduce “StreamAV-Bench: A Comprehensive Benchmark for Streaming Audio-Video Generation” 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
Kaiqi Liu, Haoxuan Zeng, Jingqi Liu, Jiacong Fang, Ziqi Cai, Yunyao Mao, Henglin Liu, Yu Sheng and 2 more
StreamAV-Bench evaluates streaming audio-video generation across progressive and interactive tracks, using 320 expert-verified scenarios and 32 dimensions across 13 systems. The interactive track measures update achievement and response latency, alongside retention and history reuse. HappyOyster achieves 0.518 visual updates, while SWIFT has 10.340-second visual latency and 0.246 achievement; latency covers achieved targets only, not all attempts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.