Metraipolis › archive › dc_ebaf9a7e58f2
Context Compaction for World Models — arXiv:2610.38980
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.38980
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T06:09:08.026Z
- SHA-256 of the archived text
36f3c15b83a4562f084c675debd2eedd69de32b8a70a1873e2ae5c354044d220- Stable URL of this page
https://metraipolis.com/documento/dc_ebaf9a7e58f2
How to cite
Context Compaction for World Models — arXiv:2610.38980 — arXiv (en), https://arxiv.org/abs/2610.38980 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ebaf9a7e58f2 (Metraipolis archive dc_ebaf9a7e58f2, sha256 36f3c15b83a4562f084c675debd2eedd69de32b8a70a1873e2ae5c354044d220)
BibTeX
@misc{dcebaf9a7e58f2,
title = {Context Compaction for World Models — arXiv:2610.38980},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.38980},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ebaf9a7e58f2; sha256 36f3c15b83a4562f084c675debd2eedd69de32b8a70a1873e2ae5c354044d220},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.38980. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
World models let observers explore and interact with generated worlds. Recent work shows that when observers revisit a region, current models often fail to regenerate it consistently. We view this as a context retrieval problem. LLM agents manage context retrieval at two layers of their stack. At the model layer, the KV cache design determines how much context fits in memory. At the agent harness layer, harnesses such as Claude Code pick the context sent to the LLM. Prior work on world models has mostly focused only on the model layer. In this paper, we make three contributions. First, we build a benchmark to measure world model context retrieval. It generates new worlds for every episode, preventing a model from recalling rooms using training data. It also separates context retrieval from guesses based on nearby frames. Second, we build an agent harness that performs context compaction for world models, which to our knowledge no prior work has studied. Third, we use our benchmark to study four world model designs. Context compaction lets our world models retrieve rooms up to 4 times beyond their context window without retraining. However, world models with sliding-window attention lose most of their context retrieval when dreaming. Therefore, a dreaming world agent benefits more from extending the reach of its world model using context compaction than switching to sliding window attention at the model layer. 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/2610.38980 — 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 2026-10-31T06:07:08.500Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).