{"type":"post","id":"memory-as-a-controlled-process-learned-adaptive-memory-4175d4","slug":"memory-as-a-controlled-process-learned-adaptive-memory-4175d4","title":"Memory as a Controlled Process: Learned Adaptive Memory Management for LLM Agents","description":"Researchers introduce MemCon, a controller that learns how an LLM agent should use an existing memory system as tasks accumulate.","retrieval_nugget":"Researchers introduce MemCon, a controller that learns how an LLM agent should use an existing memory system as tasks accumulate. Across six benchmarks, three agent frameworks and three model backbones, the paper reports up to 15.2 percentage points higher task success than multiple memory baselines, alongside 5–20% lower token use.","published_at":"2026-11-20","updated_at":"2026-11-20","record_date":"2026-11-20","date_kind":"published_at","topics":["claude"],"entities":["MemCon","arXiv","University of California, Los Angeles","University of Washington","Northwestern University","GPT-4.1-mini","Claude Sonnet 4","DeepSeek-V3.2"],"related_patterns":["company-memory-needs-write-loops","agent-security-moves-to-runtime-boundaries"],"novelty_label":"new_fact","what_changed":"MemCon introduces a learned, backend-agnostic memory policy that improves agent task success by up to 15.2 points while cutting token use by 5–20%.","prior_refs":["https://newruntime.com/posts/memory-as-middleware-for-self-improving-ai-agents-abd8b2/"],"source_url":"https://arxiv.org/abs/2607.13591","source_urls":["https://arxiv.org/abs/2607.13591"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:memory-as-a-controlled-process-learned-adaptive-memory-4175d4","status":"published","visuals":[],"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/memory-as-a-controlled-process-learned-adaptive-memory-4175d4/","markdown":"https://newruntime.com/posts/memory-as-a-controlled-process-learned-adaptive-memory-4175d4.md","json":"https://newruntime.com/posts/memory-as-a-controlled-process-learned-adaptive-memory-4175d4.json"},"markdown_url":"https://newruntime.com/posts/memory-as-a-controlled-process-learned-adaptive-memory-4175d4.md","next_reads":[{"type":"topic","path":"/topics/claude/","url":"https://newruntime.com/topics/claude/","title":"Claude","reason":"Explore the claude topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/claude.md","json_url":"https://newruntime.com/topics/claude.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/sources/","url":"https://newruntime.com/sources/","title":"Sources","reason":"See the original sources this archive cites.","media_type":"text/html"}]}
