When Agents Disagree: The Selection Bottleneck in Multi-Agent LLM Pipelines

TL;DR

This paper investigates the impact of diversity and selection methods in multi-agent LLM pipelines, revealing a crossover threshold that determines when diversity improves or worsens output quality, with judge-based selection outperforming synthesis-based methods.

Contribution

The paper introduces a theoretical crossover threshold model and provides empirical evidence showing judge-based selection surpasses synthesis in multi-agent LLM pipelines.

Findings

Diverse teams with judge-based selection outperform single models significantly.

Synthesis-based aggregation is less effective than judge-based selection across tasks.

Including weaker models can enhance performance and reduce costs.

Abstract

Multi-agent LLM pipelines produce contradictory evidence on whether team diversity improves output quality: heterogeneous Mixture-of-Agents teams outperform single models, yet homogeneous Self-MoA teams consistently win under synthesis-based aggregation. We propose a resolution by identifying the selection bottleneck -- a crossover threshold in aggregation quality that determines whether diversity helps or hurts. Under this model, we obtain a closed-form crossover threshold $s^*$ (Proposition 1) that separates the regimes where diversity helps and hurts. In a targeted experiment spanning 42 tasks across 7 categories ($N=210$), a diverse team with judge-based selection achieves a win rate of 0.810 against a single-model baseline, while a homogeneous team scores 0.512 -- near chance (Glass's $\Delta = 2.07$). Judge-based selection outperforms MoA-style synthesis by $\Delta_{\mathrm{WR}} = +0.631$ -- the synthesis approach is preferred over the baseline in zero of 42 tasks by the judge panel. A decoupled evaluation with independent judges confirms all directional findings (Spearman $\rho = 0.90$). Exploratory evidence suggests that including a weaker model improves performance while reducing cost ($p < 10^{-4}$, not pre-registered). Our results suggest that selector quality may be a more impactful design lever than generator diversity in single-round generate-then-select pipelines.