Metraipolis › archive › dc_863eecdb77e2
From Collaboration to Capability: Internalizing Routed LLM Experts into Compact Reasoners — arXiv:2609.12578
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.12578
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T03:02:49.561Z
- SHA-256 of the archived text
bf08065d9d4e68a63cd530dd6a3e5a511011dd81c4cf78f2ebe8291ccd84c5f6- Stable URL of this page
https://metraipolis.com/documento/dc_863eecdb77e2
How to cite
From Collaboration to Capability: Internalizing Routed LLM Experts into Compact Reasoners — arXiv:2609.12578 — arXiv (en), https://arxiv.org/abs/2609.12578 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_863eecdb77e2 (Metraipolis archive dc_863eecdb77e2, sha256 bf08065d9d4e68a63cd530dd6a3e5a511011dd81c4cf78f2ebe8291ccd84c5f6)
BibTeX
@misc{dc863eecdb77e2,
title = {From Collaboration to Capability: Internalizing Routed LLM Experts into Compact Reasoners — arXiv:2609.12578},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.12578},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_863eecdb77e2; sha256 bf08065d9d4e68a63cd530dd6a3e5a511011dd81c4cf78f2ebe8291ccd84c5f6},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.12578. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
A compact controller can coordinate stronger experts by selecting whom to consult, formulating requests, and integrating their responses. We study whether learning from both the controller's decisions and the experts' reasoning and code improves its generation after expert removal. We introduce \textsc{Rivet} for \emph{collaboration internalization}: expert-augmented reinforcement learning applies a shared outcome signal to controller decisions and returned expert spans, and verified trajectory internalization consolidates complete successful interactions through format-aware supervised training. The deployed controller generates reasoning, code, and interaction structure with local Python execution and no external LLM. Across seven competition-mathematics benchmarks, RIVET-1.7B and RIVET-4B achieve average accuracies of $28.25\%$ and $44.16\%$; Stage~II improves RIVET-4B's accuracy after expert removal by $6.49$ points, and GPQA-Diamond results provide evidence of generalization to scientific reasoning. Ablations show gains from ordinary trajectory supervision and additional format weighting, supporting the effectiveness of training on the content and structure of verified collaborations.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-14 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.12578 — 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 2026-10-16T02:51:27.804Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-14?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).