Metraipolis › archive › dc_c59a8eca8040

You don't need long context — arXiv:2611.05306

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.05306
License (as archived)
CC0-1.0
Archived
2026-11-04T18:35:30.309Z
SHA-256 of the archived text
9397720ebb4079646fd06dc8fdfeb7af0a4a312dfa26211a09be7dbe41b47382
Stable URL of this page
https://metraipolis.com/documento/dc_c59a8eca8040

How to cite

You don't need long context — arXiv:2611.05306 — arXiv (en), https://arxiv.org/abs/2611.05306 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c59a8eca8040 (Metraipolis archive dc_c59a8eca8040, sha256 9397720ebb4079646fd06dc8fdfeb7af0a4a312dfa26211a09be7dbe41b47382)
BibTeX
@misc{dcc59a8eca8040,
  title = {You don't need long context — arXiv:2611.05306},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.05306},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c59a8eca8040; sha256 9397720ebb4079646fd06dc8fdfeb7af0a4a312dfa26211a09be7dbe41b47382},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.05306. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Later tokens are costlier to generate than earlier tokens. On frontier open weight models, the disparity in cost may rise to as much as 7x. Through frequent compaction, we show across a broad range of long-horizon and long-input tasks that frontier language models sacrifice little to no task accuracy when maintaining short contexts. We further demonstrate that LLM performance does not collapse on long-horizon software engineering tasks even when taking at most one action per context, and based on that we outline the concept of semantic state space models (SSSMs). Our findings question the need for long-context language models.

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/2611.05306 — 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 2026-11-04T18:32:37.362Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).