Abstract
Large Language Model (LLM) based Multi-Agent Debate (MAD) is one of the most effective test time scaling techniques. Through multi-round communication, agents complement each other in knowledge and reasoning and solve tasks that no single member can solve. However, existing MAD frameworks fail to beat strong Single Agent and Consistency-based baselines under the same strict cost limit, which shakes the foundation of the MAD field. We propose Conditional Progressive Pruning (CPP), a lightweight pruning framework that fully exploits multi-round MAD. CPP outperforms all existing MAD frameworks on multiple dominated benchmarks. It is also the first to fully outperform consistency methods. Our code, detailed agent interaction records will be released soon.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Ye, R., Zhang, C., Li, J., Wu, H., Luo, X., Chen, H., Wang, Y., Chen, Y., Wang, Z., Mei, K., Zhou, Y., & Metaxas, D. N. (2026). Beyond Solo and Consistency: Vindicating Multi-Agent Debate via Conditional Progressive Pruning. https://omanscience.com/en/articles/beyond-solo-and-consistency-vindicating-multi-agent-debate-via-conditional-progressive-pruning
MLA 9
Ye, Ruosong, et al. "Beyond Solo and Consistency: Vindicating Multi-Agent Debate via Conditional Progressive Pruning." https://omanscience.com/en/articles/beyond-solo-and-consistency-vindicating-multi-agent-debate-via-conditional-progressive-pruning.
Chicago (author–date)
Ye, Ruosong, Caiqi Zhang, Jiahao Li, Haijun Wu, Xiaolong Luo, Huiyuan Chen, Yu Wang, Ying Chen, Zhenting Wang, Kai Mei, Yang Zhou, and Dimitris N. Metaxas. 2026. "Beyond Solo and Consistency: Vindicating Multi-Agent Debate via Conditional Progressive Pruning." https://omanscience.com/en/articles/beyond-solo-and-consistency-vindicating-multi-agent-debate-via-conditional-progressive-pruning.
Harvard
Ye, R., Zhang, C., Li, J., Wu, H., Luo, X., Chen, H., Wang, Y., Chen, Y., Wang, Z., Mei, K., Zhou, Y. and Metaxas, D. N. (2026) 'Beyond Solo and Consistency: Vindicating Multi-Agent Debate via Conditional Progressive Pruning', Available at: https://omanscience.com/en/articles/beyond-solo-and-consistency-vindicating-multi-agent-debate-via-conditional-progressive-pruning.
Vancouver
Ye R, Zhang C, Li J, Wu H, Luo X, Chen H, et al. Beyond Solo and Consistency: Vindicating Multi-Agent Debate via Conditional Progressive Pruning. https://omanscience.com/en/articles/beyond-solo-and-consistency-vindicating-multi-agent-debate-via-conditional-progressive-pruning
IEEE
R. Ye, C. Zhang, J. Li, H. Wu, X. Luo, H. Chen, Y. Wang, Y. Chen, Z. Wang, K. Mei, Y. Zhou, and D. N. Metaxas, "Beyond Solo and Consistency: Vindicating Multi-Agent Debate via Conditional Progressive Pruning," https://omanscience.com/en/articles/beyond-solo-and-consistency-vindicating-multi-agent-debate-via-conditional-progressive-pruning.