{"version":1,"type":"story","url":"https://digestai.news/story/researchers-propose-fem-asm-to-separate-storage-execution-and-coordina","json":"https://digestai.news/story/researchers-propose-fem-asm-to-separate-storage-execution-and-coordina.json","markdown":"https://digestai.news/story/researchers-propose-fem-asm-to-separate-storage-execution-and-coordina.md","slug":"researchers-propose-fem-asm-to-separate-storage-execution-and-coordina","headline":"Researchers propose FEM-ASM to separate storage, execution, and coordination in language models","summary":"The article introduces FEM-ASM, a finite-element-method-inspired organization for language models that separates contextual computation, persistent storage, and exact execution. It uses independently constructed document states and deterministic executable skills that contribute typed proposals to a shared model state, reconciled by an explicit residual operator. The approach is evaluated through controlled experiments and negative results, without claiming a physical finite-element formulation. A versioned store contains 52,809 reconstructive memory elements near a 1.7-billion-floating-value budget, with reconstruction showing approximately 75% token accuracy. Support-aware lexical indices make these elements addressable under provenance-controlled queries. For executable arithmetic, positional result observations improve neural rendering over a repeated global result vector, and output substitutions change the model's preferred answer. A bounded attachment demonstration measures the effect of making selected evidence available, without claiming utility for loading an entire multi-billion-value store. The results support separating storage, execution, and neural coordination while identifying limitations in question-only retrieval, unrestricted answer generation, and end-to-end efficiency.","keyPoints":["FEM-ASM separates document states, executable skills, and a shared model state with a residual operator","Versioned store has 52,809 reconstructive memory elements near a 1.7-billion-floating-value budget","Reconstruction achieves approximately 75% token accuracy; positional observations improve neural rendering"],"whyItMatters":"The work explores architectural alternatives to monolithic parameter updates in language models, offering a path toward more modular and interpretable systems, though practical capabilities remain limited.","category":{"slug":"research","name":"Research","url":"https://digestai.news/category/research"},"entities":{"companies":[],"models":["FEM-ASM"],"people":[]},"firstPublishedAt":"2026-10-06T04:00:00Z","updatedAt":"2026-10-06T04:00:00Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"arXiv cs.AI","title":"Beyond the Parameter Monolith: Reconstructive Memories, Executable Skills, and Residual Assembly for Language Models","url":"https://arxiv.org/abs/2610.04012","publishedAt":"2026-10-06T04:00:00Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Researchers propose FEM-ASM to separate storage, execution, and coordination in language models\", 6 October 2026, https://digestai.news/story/researchers-propose-fem-asm-to-separate-storage-execution-and-coordina","publisher":"Digest AI","title":"Researchers propose FEM-ASM to separate storage, execution, and coordination in language models","datePublished":"2026-10-06T04:00:00Z","url":"https://digestai.news/story/researchers-propose-fem-asm-to-separate-storage-execution-and-coordina"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}