Reproduce “When Order Matters: First-Speaker Bias and Mitigation through Personality in Sequential Multi-Agent Debate” 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
Duofeng Xu, Bryan Hooi, Dandan Qiao
The authors test how sequential speaking order affects multi-agent debate, using two weaker agents and one stronger agent across four multiple-choice benchmarks. Moving the stronger agent from last to first raises its influence by 21.01 percentage points and final-answer accuracy by 2.34 points. These effects include resulting changes in dialogue and do not identify a psychological mechanism.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.