StagedWorkspace: A Versioned Workspace for Knowledge-Work Agents
2026-08-18 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors explain that AI systems working on tasks like coding or editing documents often handle different versions of files, which can cause confusion. They propose StagedWorkspace, a system that keeps track of these versions by linking views and edits to specific file states. Their tests show that having access to both the original files and their parsed versions helps AI agents perform better on tasks like answering questions about documents and editing files. The work highlights the importance of managing workspace state to improve AI knowledge work.
AI agentsknowledge workworkspace stateversion controlparsed viewsdiffsdigital artifactsOfficeQAAPEX-Agentscontent hashes
Authors
Yining Hua, Hongbin Na, Yifan Zhou, Akshay Kalose, Cyrus Ayubcha, Levi Lian
Abstract
AI agents increasingly perform knowledge work (i.e., produce and modify persistent digital artifacts such as code repositories, documents, spreadsheets, slides, reports), yet the parsed views they search, the native files they edit, the changes they review, and the artifacts they submit can refer to different versions of the same work product. We formulate this as a workspace-state contract: every view should be explicitly tied to a version of the evolving workspace state. Coding agents partly address this need through repository contracts for search, diffs, and tests, whereas an analogous contract is less explicit for PDFs, spreadsheets, slides, notebooks, and mixed-format project folders. We propose StagedWorkspace, a versioned workspace for knowledge-work agents. The workspace binds parsed records and review diffs to content hashes of the native files as they change. In fixed-harness ablations on OfficeQA Pro and APEX-Agents, dual parsed/native access has the highest point estimate for every tested model; relative to the more limiting single view, it improves OfficeQA Pass@1 by 8.3-12.1 points and APEX mean rubric score by 4.7-9.2 points. SW-AGENT scores 63.9% with Gemini 3.1 Pro on OfficeQA and 42.1 with GPT-5.4 Nano on APEX, compared with published same-model scores of 29.3% and 25.5, respectively. A paired review-axis ablation on 57 file-editing tasks further finds higher observed scores when diffs are visible. These results identify workspace state as an experimental variable in knowledge-work agents and motivate benchmarks that score evidence, staged edits, and submitted artifacts as explicit state transitions.