Metraipolis › archive › dc_015389fceca9
Think More or Read More? A Token-Budget Theory of Reasoning with Shared Context — arXiv:2610.23246
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.23246
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T08:43:48.037Z
- SHA-256 of the archived text
67d5e2e27c188e2798a345678d66d1655c1c3fe3000aeccbf20cea632575ebde- Stable URL of this page
https://metraipolis.com/documento/dc_015389fceca9
How to cite
Think More or Read More? A Token-Budget Theory of Reasoning with Shared Context — arXiv:2610.23246 — arXiv (en), https://arxiv.org/abs/2610.23246 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_015389fceca9 (Metraipolis archive dc_015389fceca9, sha256 67d5e2e27c188e2798a345678d66d1655c1c3fe3000aeccbf20cea632575ebde)
BibTeX
@misc{dc015389fceca9,
title = {Think More or Read More? A Token-Budget Theory of Reasoning with Shared Context — arXiv:2610.23246},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.23246},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_015389fceca9; sha256 67d5e2e27c188e2798a345678d66d1655c1c3fe3000aeccbf20cea632575ebde},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.23246. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Reasoning systems increasingly improve answers by incorporating additional context into the model's input, from multi-agent debate, where agents read their peers' responses, to self-refinement, where a model revisits its own drafts. In these systems, communication consumes context as one agent's output becomes another's input. We study the regime where token reading and generation draw on one fixed token window, so every token spent reading is a token that cannot be spent reasoning. We develop a \textbf{token-budget theory} that formalizes this trade-off, modeling accuracy as a function of the reasoning budget remaining after reading and aggregating votes with a Condorcet jury model that permits correlated errors. For multi-agent debate, the theory gives a closed-form prediction for the optimal number of agents. This quantity is estimated from single-agent measurements, allowing the optimal group size to be determined without running debate, and the same accounting extends to a heterogeneous set of different models. We empirically verify our theory and show that, at the predicted group size, our method outperforms strong baselines under the tightest calibration spend, where debate-calibrated rules can afford only one or two questions. Self-refinement obeys the same accounting, with its own draft in place of a peer message. The theory further predicts that communication pays only when the gain in the accuracy ceiling exceeds the reasoning room spent on reading, with this condition tightening as the window shrinks. Across five open models and four reasoning tasks, self-consistency matches or beats deployed debate at $93$\% of $980$ matched window-and-group-size settings, with the gap being largest under tight windows and nearly disappearing at a loose one, consistent with the theory. Under a shared window, how much room a model has left to \textit{think matters more than what it reads}.
Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.23246 — 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 2026-10-19T08:39:24.864Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).