Beyond Consensus: Trace-Level Synthesis in Mixture of Agents
For practitioners using LLM ensembles, this work demonstrates that trace-level synthesis provides a more effective aggregation method than consensus-based approaches, with potential to improve reasoning in multi-agent systems.
The paper shows that aggregating full reasoning traces from LLM agents, rather than just their final answers, yields better performance than majority voting, even when agents unanimously agree. The proposed Self-Consistent Mixture of Agents method outperforms heterogeneous model pools on multiple benchmarks.
When multiple LLM agents solve the same problem, standard practice compresses each agent's reasoning into a majority vote or layered synthesis, treating agreement as the finish line. We show this is unnecessarily lossy: an LLM aggregator that reads complete reasoning traces recovers correct solutions even when agents unanimously agree, with beneficial corrections consistently outweighing harmful ones -- the \emph{aggregation paradox}. Majority voting has a ceiling that perturbation diversity does not raise (error correlations are identical); the aggregator's gain comes from trace-level complementarity, assembling correct intermediate steps from minority chains that voting discards. These findings motivate Self-Consistent Mixture of Agents which generates trace diversity through semantic-preserving input perturbations, safeguards the majority via anchored refinement with provable non-degradation guarantees, and always synthesizes -- never gates on consensus. A single model with perturbation-induced trace variation outperforms heterogeneous model pools across structured reasoning, PhD-level science, competition mathematics, and competitive programming. The unit of aggregation should be the reasoning trace, not the answer.