PaperJury: Due-Process Review for Bounded LaTeX Revision

Researchers have introduced PaperJury, a closed-loop system for revising LaTeX computer science papers that uses a deterministic orchestration layer to manage review, verdict, and revision while restricting semantic models to bounded judgment and repair tasks [1]. The system, detailed in a paper submitted to arXiv on June 15, 2026, addresses what its creators describe as a gap in existing writing assistants: the absence of durable issue identity across revision rounds, deterministic routing from critique to adjudication, and the ability to reject invalid concerns or defer author-dependent ones [1]. PaperJury operates on a deterministic-versus-semantic split. A deterministic orchestration layer handles decomposition, a frozen claim spine, a durable ledger, routing, stopping, and exact-once patch application. Semantic agents are confined to bounded review, judgment, and repair [1]. The architecture combines bounded holistic review, contestability-based routing, a due-process trial, and risk-proportional guard chains for anchor-bounded edits. Each review cycle produces one of three terminal outcomes: invalid-drop, valid-fixable, or author-required [1]. The system is available at a public GitHub repository [2]. The researchers evaluated PaperJury in a two-arm expert-review study using held-out papers from the Vision, natural language processing, and machine learning domains, comparing it against four baselines [1]. The evaluation measured issue quality, verdict and routing quality, edit safety, convergence behavior, and cost [2]. The results support the thesis that load-bearing safety and completion logic should reside in deterministic orchestration rather than model discretion [1]. The work builds on a broader recognition that automated writing tools for academic manuscripts require different safeguards than general-purpose drafting assistants. Pre-submission hardening demands adversarial whole-paper review, explicit no-fix outcomes, and bounded artifact-safe revision, requirements that critique generators and judge-centered loops have not fully met [1]. The deterministic orchestration approach in PaperJury represents a structural response to those demands, separating the reliability-critical control flow from the probabilistic outputs of language models [2]. The paper appears on arXiv under the Computation and Language category and is associated with arXivLabs, a framework that allows collaborators to develop and share new arXiv features directly on the platform [1].