Reproduce “DHCG: Dynamic Construction of Hierarchical Collaboration Graphs for LLM-Based Multi-Agent Reasoning” 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
Jie Ren, Jiakang Yuan, Chenyu Huang, Hezeer Ma, Jiayuan Fan, Tao Chen
DHCG dynamically constructs a team of language-model agents by creating initial roles, then using the latest workers’ outputs to add roles with selected inputs or stop and send selected outputs to a Generator. Across seven benchmarks on Qwen3-8B, it averages 73.43%, above IO, G-Designer, and MaAS, but does not lead every benchmark: LLM-Debate scores higher on GSM8K.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.