Reproduce “R²-Omni: Reducing Structural and Precision Redundancy in Omni-LLMs for Long-Form 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
Bolong Feng, Pengfei Zhou, Zhenglin Wan, Wei Wang, Wangbo Zhao
R²-Omni compresses long audio-video caches for open-ended descriptions by merging locally repetitive audio and video tokens, then assigning retained cache values different numerical precisions. It reports average 5.0× KV-cache compression while retaining 97.98% of baseline quality across three Omni-LLMs and three long-form benchmarks. Separately, it achieves 1.48× end-to-end speedup for Qwen2.5-Omni-7B at 256K context, not established for other context lengths.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.