Metraipolis › archive › dc_5e4fd16a68df

CTXAgent - Self-Controlled Context Management of LLM Agents — arXiv:2610.34425

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.34425
License (as archived)
CC0-1.0
Archived
2026-10-27T07:57:31.202Z
SHA-256 of the archived text
063668a09d67fb2e551b338f6d507ac56d8db71881b8eae93bea17be6ccf0f41
Stable URL of this page
https://metraipolis.com/documento/dc_5e4fd16a68df

How to cite

CTXAgent - Self-Controlled Context Management of LLM Agents — arXiv:2610.34425 — arXiv (en), https://arxiv.org/abs/2610.34425 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_5e4fd16a68df (Metraipolis archive dc_5e4fd16a68df, sha256 063668a09d67fb2e551b338f6d507ac56d8db71881b8eae93bea17be6ccf0f41)
BibTeX
@misc{dc5e4fd16a68df,
  title = {CTXAgent - Self-Controlled Context Management of LLM Agents — arXiv:2610.34425},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.34425},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_5e4fd16a68df; sha256 063668a09d67fb2e551b338f6d507ac56d8db71881b8eae93bea17be6ccf0f41},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.34425. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Agent harnesses keep long sessions inside the context window by compacting the history when a token threshold is crossed. We study an alternative in which the model manages its own context: it decides when to compact, what to keep and what to write as the hand-off, given two tools and one paragraph of guidance. Since earlier work found that open-weight models cannot manage their own context without scaffolding, we compare seven frontier models under identical conditions, with and without the context tools, on token cost and benchmark score, and we compare our harness with three widely used harnesses on two of these models. The benchmark is assembled from existing ones: coding exercises, multi-step SciCode problems and SWE-bench Pro and Terminal-Bench 4.0 tasks are interleaved with planted facts, ephemeral observations and standing instructions, all checked deterministically, in sessions of 35-212 turns that resemble a user working through different tasks in one chat. Most of the strongest models manage their context: they compact and curate regularly, the live context of the mixed sessions stays mostly below 100k tokens, and their benchmark scores mostly do not drop significantly. For the models that manage their context well, the cost reduction is substantial but varies between models, because some models compact so often that other costs increase. With Claude Opus 5.5, the self-managed harness matches Claude Code on task success at 34% lower cost than its explicit 200k-window configuration and 49% lower than its default, with no context window to choose. We release the agent, the benchmark and all result rows.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.34425 — 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 2026-10-27T07:55:30.686Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).