Metraipolis › archive › dc_f8f7463ba2e5

One Model, Many Minds: Unlocking Multi-Agent Synergy in a Single Agent via Mixture of Roles — arXiv:2608.27338

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.27338
License (as archived)
CC0-1.0
Archived
2026-10-17T13:16:36.946Z
SHA-256 of the archived text
3e9539e66bdfcbc880d2364e60350a3d8d6e33b62d4b42a8dc211da66d9e39d5
Stable URL of this page
https://metraipolis.com/documento/dc_f8f7463ba2e5

How to cite

One Model, Many Minds: Unlocking Multi-Agent Synergy in a Single Agent via Mixture of Roles — arXiv:2608.27338 — arXiv (en), https://arxiv.org/abs/2608.27338 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f8f7463ba2e5 (Metraipolis archive dc_f8f7463ba2e5, sha256 3e9539e66bdfcbc880d2364e60350a3d8d6e33b62d4b42a8dc211da66d9e39d5)
BibTeX
@misc{dcf8f7463ba2e5,
  title = {One Model, Many Minds: Unlocking Multi-Agent Synergy in a Single Agent via Mixture of Roles — arXiv:2608.27338},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.27338},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f8f7463ba2e5; sha256 3e9539e66bdfcbc880d2364e60350a3d8d6e33b62d4b42a8dc211da66d9e39d5},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.27338. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Specializing Large Language Models (LLMs) toward distinct abilities underpins successes ranging from personalized assistants to multi-agent systems (MAS). Single-agent paradigms rely on pre-defined personas or steering vectors to induce specialization, yet they impose a single fixed specialization that fails to adapt to diverse queries. Conversely, MAS achieves dynamic multi-perspective problem solving by orchestrating agents with distinct text-based roles, but fusing these specializations requires multi-turn interactions that inflate context length and inference cost. To address these limitations, we propose Mixture of Roles (MoRe), which adaptively composes multiple specializations into a single steering vector for single-turn inference. Specifically, MoRe learns a diversified codeboox of steering vectors, each of which encodes a latent role. A query-aware router dynamically fuses the codebook into a steering vector that encompasses multiple roles. By steering the backbone LLM with the composed vector, MoRe enables multi-perspective specialization in a single-agent, single-turn inference process. The proposed MoRe can be efficiently trained via a three-stage SFT curriculum and GRPO post-training, while the backbone LLM remains frozen. Experiments across reasoning and personality benchmarks show that MoRe outperforms single-agent baselines by 2.2% on average, and achieves performance on par with MAS while reducing token cost by 20x.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-28 in cs.MA).
Categorie arXiv: Multiagent Systems (cs.MA).
Pagina del lavoro: https://arxiv.org/abs/2608.27338 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T13:16:16.994Z dalla pagina https://arxiv.org/catchup/cs.MA/2026-08-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).