Metraipolis › archive › dc_063df4a31db2
When Evidence Changes: Evaluating Memory Repair and Re-reading in Language-Model Agents — arXiv:2610.03902
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.03902
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T05:27:58.471Z
- SHA-256 of the archived text
cba200ee3a83f969a8379a873c6f4aa9057e0862f28d54263fecf378e0647abb- Stable URL of this page
https://metraipolis.com/documento/dc_063df4a31db2
How to cite
When Evidence Changes: Evaluating Memory Repair and Re-reading in Language-Model Agents — arXiv:2610.03902 — arXiv, https://arxiv.org/abs/2610.03902 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_063df4a31db2 (Metraipolis archive dc_063df4a31db2, sha256 cba200ee3a83f969a8379a873c6f4aa9057e0862f28d54263fecf378e0647abb)
BibTeX
@misc{dc063df4a31db2,
title = {When Evidence Changes: Evaluating Memory Repair and Re-reading in Language-Model Agents — arXiv:2610.03902},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.03902},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_063df4a31db2; sha256 cba200ee3a83f969a8379a873c6f4aa9057e0862f28d54263fecf378e0647abb},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.03902. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
When documents supporting an agent's derived facts are revoked or replaced, should it repair memory or re-read current evidence? We introduce an evidence-revision evaluation on medication- and problem-list tasks from public ICU records. Under revocation, replacement and control events, we compare full and source-filtered re-reading with caching, rebuilding and graph-local repair across two 7B models. Memory is supplied in full without retrieval, and costs include ingest, revision and every use. On short records, local repair uses 5-10$\times$ fewer revision tokens than rebuilding, yet every memory pipeline costs at least twice full re-reading in held-out conditions. In a small pre-specified development sweep, adding task-ineligible documents extended records to about 10,000 tokens; at that length, memory's mean cumulative cost fell below full re-reading's after 2-14 uses, partly through truncated extraction, while source-filtered re-reading remained cheapest. In the replacement study, none of the four primary confirmatory tests reached statistical significance. These results show why the cost of agent memory after evidence revision must be assessed against source-filtered re-reading over the full pipeline. 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/2610.03902 — 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 2026-10-06T05:24:20.239Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).