Reproduce “AIR-LLM: Broadcasting AI Weights over Radio for Memory-Free Edge LLM Inference via RF Computing” 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
Zhihui Gao, Tingjun Chen, Dirk Englund
AIR-LLM evaluates edge language-model inference using broadcast weights, avoiding their storage on devices. Each device combines arriving weights with local activations in an RF mixer. On WikiText-2, a selected Paris user gets perplexity 6.49 versus 6.24 for FP16, a reported 4.0% degradation. Estimated edge energy per token is 157.7× lower than remote FP16; city channels are simulated and mixer distortion separately measured.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.