Abstract page for arXiv paper 2610.35957: Distributed LLM Inference via Communication-Efficient Speculative Decoding