{"slug": "mom-memory-of-memory", "title": "MoM: Memory of Memory", "summary": "Researchers introduced Memory of Memory (MoM), a memory design for long-horizon LLM agents that commits a current value on arrival while retaining displaced values as provenance, instantiated as Provenant Memory (P-Mem), a typed provenance graph. In the arXiv paper 2609.25054v1, P-Mem matched the strongest retrieval memory in turn-level read accuracy using roughly 4x fewer read tokens, cut the knowledge-update stale-answer rate from 19.4% to 10.9% via graph-guided turn pruning, held 100% accuracy on revision chains where query-time reading collapsed to 25%, and recovered committed errors at 100% versus 0% for a CRUD memory.", "body_md": "arXiv:2609.25054v1 Announce Type: new \nAbstract: For a long-horizon LLM agent, the memory question is not what was once recorded but what \\emph{currently holds}. Most designs answer it only indirectly: every interaction is stored, and the present is reconstructed at query time by retrieving and reconciling records, so stale values re-enter and the same conflicts are re-litigated. Committing the current value at write time avoids this, but existing write-time (CRUD) memories overwrite, so a wrong update is unrecoverable and prior state is lost. We take the missing combination---\\emph{commit on arrival while retaining what is displaced}---and formalize it as \\textsc{Memory of Memory} (MoM): memory tracks not only content but the provenance, status, and history of its own entries. We instantiate MoM as \\textsc{Provenant Memory} (P-Mem), a typed provenance graph whose \\emph{active frontier} exposes one current value per resolved key while displaced values are retained as provenance; typed operations decide whether a new observation supports, supersedes, contests, rejects, revokes, or resolves an existing value. P-Mem's decisive gain is validity rather than accuracy: its turn-level read matches the strongest retrieval memory in accuracy at $\\sim$4$\\times$ fewer read tokens---a retrieval-granularity effect---while graph-guided turn pruning cuts the knowledge-update stale-answer rate (19.4\\%$\\rightarrow$10.9\\%); on revision chains it stays at 100\\% where query-time reading collapses to 25\\%, and, because displaced values are retained rather than overwritten, it recovers committed errors a CRUD memory cannot (100\\% vs.\\ 0\\%).", "url": "https://wpnews.pro/news/mom-memory-of-memory", "canonical_source": "https://arxiv.org/abs/2609.25054", "published_at": "2026-09-23 04:00:00+00:00", "updated_at": "2026-09-23 04:25:38.736094+00:00", "lang": "en", "topics": ["ai-agents", "large-language-models", "ai-research", "artificial-intelligence"], "entities": ["Memory of Memory", "Provenant Memory", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/mom-memory-of-memory", "markdown": "https://wpnews.pro/news/mom-memory-of-memory.md", "text": "https://wpnews.pro/news/mom-memory-of-memory.txt", "jsonld": "https://wpnews.pro/news/mom-memory-of-memory.jsonld"}}