Metraipolis › archive › dc_65a80b4fc741
From Long Context to Selective Memory:A Survey of Memory-Centric Compression for Large Language Models — arXiv:2610.21165
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.21165
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T09:55:33.188Z
- SHA-256 of the archived text
a40a8728f8189be127fa4d887d8caad06a56783bd9ce738c1e6d1ec43abf8dbf- Stable URL of this page
https://metraipolis.com/documento/dc_65a80b4fc741
How to cite
From Long Context to Selective Memory:A Survey of Memory-Centric Compression for Large Language Models — arXiv:2610.21165 — arXiv (en), https://arxiv.org/abs/2610.21165 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_65a80b4fc741 (Metraipolis archive dc_65a80b4fc741, sha256 a40a8728f8189be127fa4d887d8caad06a56783bd9ce738c1e6d1ec43abf8dbf)
BibTeX
@misc{dc65a80b4fc741,
title = {From Long Context to Selective Memory:A Survey of Memory-Centric Compression for Large Language Models — arXiv:2610.21165},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.21165},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_65a80b4fc741; sha256 a40a8728f8189be127fa4d887d8caad06a56783bd9ce738c1e6d1ec43abf8dbf},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21165. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Long-context capability has become essential for large language models (LLMs) to handle complex tasks, yet continuously expanding the context window does not remove the computational, storage, and information-selection pressures incurred at inference time. Existing work on prompt compression, KV-cache management, efficient serving, retrieval augmentation, and agent memory is usually discussed in isolation, leaving the flow of information between storage and computation insufficiently understood. This survey reframes long-context compression as a memory-management problem for allocating inference-time resources. We distinguish compute-active memory, which directly participates in model computation, from compute-inactive memory, which lies outside the computation path but can be stored, retrieved, updated, or forgotten, and we characterize offloaded KV caches, persistent prefix caches, and precomputed KV states as intermediate states on the activation path. On this basis, we formulate writing, offloading, recall, retrieval-and-injection, materialization, eviction, consolidation, and forgetting as state transitions within an information lifecycle, and introduce a budgeted cost model that places these operations in a single cost account. The resulting framework unifies short-term context compression, long-term memory management, and the transitions between them, and shows that the central challenge is not merely reducing information volume but deciding which information should participate in costly inference, when, and in what form. We further summarize evaluation dimensions, benchmark gaps, and open challenges, and argue that future research should move from static, pointwise compression toward dynamic, governable, lifecycle-aware memory management. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2610.21165 — 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 2026-10-16T08:44:47.628Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).