| title | CERA-MoA: Co-Evolving Routing Mechanisms with Continually Learning LLM Agents |
|---|---|
| authors | Jiaxuan Jiang, Liyuan He, Zhixuan Fang |
| date | 2026-09-16 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2609.18779 |
| affiliation | Tsinghua University |
| upvotes | 15 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Current Mixture-of-Agents (MoA) paradigms generally treat query routing and agent fine-tuning as separate processes, limiting their ability to respond to evolving agent capabilities. This disconnect prevents routing strategies from adapting to evolving agent capabilities during post-training and prevents agents from achieving synergistic data-driven specialization. To resolve this, we introduce CERA-MoA (Co-Evolving Router with continually learning Agents for Mixture-of-Agents), an iterative reinforcement learning framework where the dynamic router and independent agent policies co-evolve. We design a predictive familiarity estimator that leverages mid-layer hidden states to evaluate semantic competence among agents, avoiding the overhead of full rollouts. Based on these familiarity scores, a cumulative-threshold adaptive routing mechanism dynamically activates a tailored minimal agent subset, achieving a trade-off between task performance and efficiency. By proactively allocating targeted training samples to agents based on their evolving competence, CERA-MoA promotes capability differentiation. Extensive experiments across various domains demonstrate that CERA-MoA outperforms state-of-the-art static-agent routing and fix-workflow fine-tuning baselines. |
cite
@misc{jiang2026cera,
title={CERA-MoA: Co-Evolving Routing Mechanisms with Continually Learning LLM Agents},
author={Jiaxuan Jiang and Liyuan He and Zhixuan Fang},
year={2026},
eprint={2609.18779},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2609.18779}
}