{"ID":6024022,"CreatedAt":"2026-07-08T01:00:23.257252134Z","UpdatedAt":"2026-07-09T16:12:55.966383441Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.05483","arxiv_id":"2607.05483","title":"PatchOptic for Shared-State LLM Workflows with Projected Views and Verified Structured Updates","abstract":"Agentic workflows often operate over shared, structured state. Because LLM context windows are limited, each model invocation is typically shown only the state fragment needed for the current workflow step, a pattern commonly known as progressive disclosure. Modern systems construct such model-facing views using grep-like keyword search, retrieval-augmented generation (RAG), abstract-syntax-tree (AST) queries, and task-specific agent skills. These methods make the read side manageable, but they do not define when a locally proposed rewrite is valid after it is applied back to the full state. The missing piece is a contract between local updates and global validity. We introduce PatchOptic, an optic-inspired interface for shared-state LLM workflows. Optics are compositional bidirectional accessors that describe how views of structured data are read and updated. PatchOptic borrows this view/update intuition and realizes it through projected reads and verified structured patches. Each workflow step declares a projected read view, an authorized write region, and a patch-source region. Beyond runtime enforcement, the same declaration yields a path-level footprint that supports delegation, sub-workflow composition, and static certificates for reordering independent steps within the same phase. We evaluate this design with PatchBench, a benchmark with 46 cases across domains. The results show that projected reads reduce reported leakage and token cost while preserving accepted-output quality under the strong actor. Runtime verification blocks declared workflow-contract violations before commit, and patch-read enforcement rejects compromised patch artifacts that use hidden sources.","short_abstract":"Agentic workflows often operate over shared, structured state. Because LLM context windows are limited, each model invocation is typically shown only the state fragment needed for the current workflow step, a pattern commonly known as progressive disclosure. Modern systems construct such model-facing views using grep-l...","url_abs":"https://arxiv.org/abs/2607.05483","url_pdf":"https://arxiv.org/pdf/2607.05483v1","authors":"[\"Zhaoyu Bai\",\"Jiaqi Cai\"]","published":"2026-07-06T16:13:56Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.AI\",\"cs.LO\",\"cs.MA\",\"cs.PL\"]","methods":"[\"RAG\",\"Large Language Model\"]","has_code":false}