Metraipolis › archive › dc_b6daba5fc90d
Attention Tax, Handoff Tax: A Stylised Model of When Multi-Agent LLM Systems Help — arXiv:2610.06069
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.06069
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T05:23:54.901Z
- SHA-256 of the archived text
1390f09057ceb349161ca0c77081cb4b6c31d2584468a226b924366eeb22f69e- Stable URL of this page
https://metraipolis.com/documento/dc_b6daba5fc90d
How to cite
Attention Tax, Handoff Tax: A Stylised Model of When Multi-Agent LLM Systems Help — arXiv:2610.06069 — arXiv, https://arxiv.org/abs/2610.06069 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b6daba5fc90d (Metraipolis archive dc_b6daba5fc90d, sha256 1390f09057ceb349161ca0c77081cb4b6c31d2584468a226b924366eeb22f69e)
BibTeX
@misc{dcb6daba5fc90d,
title = {Attention Tax, Handoff Tax: A Stylised Model of When Multi-Agent LLM Systems Help — arXiv:2610.06069},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.06069},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b6daba5fc90d; sha256 1390f09057ceb349161ca0c77081cb4b6c31d2584468a226b924366eeb22f69e},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.06069. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Recent work on multi-agent LLM systems reaches sharply different conclusions: some results show that a single agent with the same information and compute should dominate a delegated system, others that multi-agent gains grow with task depth. We argue that much of the disagreement comes from modelling different bottlenecks, and introduce a stylised reliability model built around two trade-offs. Decomposition reduces the burden of long contexts but incurs a handoff tax when information is compressed or transferred between agents. Redundancy gains from multiple samples, but its benefit depends on how much their failures are shared. With reasoning budget, verification, and task structure added, the model yields two crossover conditions: decomposition becomes preferable once the attention cost avoided by resetting context exceeds the handoff cost, and parallel sampling at equal budget is eventually preferable when its shared-failure floor lies below the error floor of one agent thinking longer. We connect these regimes to recent theoretical and empirical results. On a ledger-reconciliation task we measure the context-degradation curve and the handoff tax from single-agent and handoff runs alone. From these the model places the crossover at depth 10 and predicts decomposition to win at depths 20, 50, and 100. It does, on step-level and final-balance accuracy, and the decomposed system's success, which the prediction never sees, lands within 9 percentage points of the predicted rate at every depth. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.MA, nuovi lavori del giorno). Categorie arXiv: Multiagent Systems (cs.MA) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.06069 — 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 2026-10-06T05:23:50.213Z dalla pagina https://arxiv.org/list/cs.MA/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).