{"ID":5552875,"CreatedAt":"2026-07-02T01:54:51.863792489Z","UpdatedAt":"2026-07-03T23:33:56.302496761Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.00269","arxiv_id":"2607.00269","title":"Mnemosyne: Agentic Transaction Processing for Validating and Repairing AI-generated Workflows","abstract":"LLMs, solvers, and agent teams increasingly generate workflow actions, repairs, and plans, but a generated action may be syntactically valid yet stale, infeasible, conflicting, or destructive of the evidence that triggered a repair. We introduce Agentic Transaction Processing (ATP), a transaction model that treats generated actions as untrusted proposals until they pass deterministic admission under a declared, executable constraint set C. The principle is two-sided: a proposal is not truth, and no proposal foresees every disruption: anything may propose, but only the runtime admits and commits, and when an unforeseen disruption strikes it repairs reactively within bounds rather than trusting a fresh proposal. Relative to C, committed-state correctness becomes independent of the competence, honesty, or learning of the proposing layer. We realize ATP in Mnemosyne, a runtime with an append-only transition log, effective-state projection, dependency-safe compensation, and active commitment records, and prove four safety properties relative to C (authority separation, serial-equivalent generative admission, evidence-preserving repair, and obligation containment) together with a bounded-reactive-repair guarantee for its localized repair protocol (LCRP). A reproducible artifact rejects the targeted violations across nine falsification tests while still admitting valid work, at under 6% projection-and-validation overhead, and bounded local repair edits an order of magnitude fewer operations than global recompute. Mnemosyne is open source: https://github.com/eyuchang/Mnemosyne/tree/arxiv-atp-rq1-rq9b-r8-v2.","short_abstract":"LLMs, solvers, and agent teams increasingly generate workflow actions, repairs, and plans, but a generated action may be syntactically valid yet stale, infeasible, conflicting, or destructive of the evidence that triggered a repair. We introduce Agentic Transaction Processing (ATP), a transaction model that treats gene...","url_abs":"https://arxiv.org/abs/2607.00269","url_pdf":"https://arxiv.org/pdf/2607.00269v1","authors":"[\"Edward Y. Chang\",\"Longling Geng\",\"Emily J. Chang\"]","published":"2026-06-30T23:33:16Z","proceeding":"cs.AI","tasks":"[\"cs.AI\"]","methods":"[\"Large Language Model\"]","has_code":false,"code_links":[{"ID":613866,"CreatedAt":"2026-07-02T01:54:51.863792489Z","UpdatedAt":"2026-07-02T01:54:51.863792489Z","DeletedAt":null,"paper_id":5552875,"paper_url":"https://arxiv.org/abs/2607.00269","paper_title":"Mnemosyne: Agentic Transaction Processing for Validating and Repairing AI-generated Workflows","repo_url":"https://github.com/eyuchang/Mnemosyne","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}