Reproduce “Experts are all you need: A Composable Framework for Large Language Model Inference” 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
Shrihari Sridharan, Sourjya Roy, Anand Raghunathan, Kaushik Roy
Comp-LLM answers complex questions by decomposing them into dependent sub-questions, routing each to pretrained domain experts, and scheduling independent steps in parallel. Its dependency graph passes prerequisite answers to later steps. On MultiExpertQA-All, the 3.65B-parameter OPT system reports 0.62 accuracy versus 0.57 for base OPT-13B. The method targets static questions; interactive tasks remain outside its effective scope.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.