PaperJury: Due-Process Review for Bounded LaTeX Revision

2026-06-15 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Research Methodology & Innovation · Depth: Expert, quick

Summary

PaperJury is a novel closed-loop review-verdict-revise-verify system designed for pre-submission hardening of human-authored LaTeX computer science papers. It addresses limitations of existing writing assistants by providing adversarial whole-paper review, explicit no-fix outcomes, and bounded artifact-safe revision. The system employs a deterministic-versus-semantic split, where deterministic orchestration manages decomposition, a frozen claim spine, a durable ledger, routing, stopping, and exact-once patch application. Semantic agents are restricted to bounded review, judgment, and repair. PaperJury integrates bounded holistic review, contestability-based routing, a due-process trial, and risk-proportional guard chains, yielding terminal outcomes of invalid-drop, valid-fixable, and author-required. An expert-review evaluation on Vision, natural language processing, and machine learning papers against four baselines supports the thesis that load-bearing safety and completion logic should reside in deterministic orchestration rather than model discretion.

Key takeaway

For research scientists and AI architects developing or reviewing technical papers, PaperJury highlights the critical importance of embedding load-bearing safety and completion logic within deterministic orchestration, rather than relying solely on AI model discretion. You should consider implementing structured, due-process review systems that ensure durable issue identity and bounded, artifact-safe revisions. This approach can significantly enhance the reliability and integrity of pre-submission hardening processes.

Key insights

PaperJury offers a robust, deterministic system for safe and auditable LaTeX paper revision.

Principles

• Deterministic orchestration ensures safety and completion logic.
• Adversarial review requires durable issue identity.
• Bounded revision prevents artifact-unsafe changes.

Method

PaperJury uses a closed-loop review-verdict-revise-verify system with a deterministic-versus-semantic split, orchestrating decomposition, routing, and exact-once patch application.

In practice

• Implement contestability-based routing for critiques.
• Define explicit terminal outcomes: invalid-drop, valid-fixable, author-required.
• Utilize guard chains for anchor-bounded edits.

Topics

• LaTeX Revision
• Deterministic Orchestration
• Adversarial Review
• Due-Process Review
• Machine Learning Papers
• NLP Papers
• Vision Papers

Code references

• u7079256/paperjury

Best for: AI Scientist, Research Scientist, AI Architect

Related on AIssential

• It's not the Language Model, it's the Tool: Deterministic Mediation for Scientific Workflows — Typed mediation uses LLMs to orchestrate deterministic, researcher-defined tools for reproducible scientific analysis, ensuring consistent results.
• ICML paper checker is down? [D] — Website outages near deadlines disrupt critical submission processes, necessitating immediate communication and contingency planning.
• Can AI Review Improve Paper Drafting? An Empirical Study on 20 Computer Architecture Submissions — AI review can significantly align with human feedback and identify novel issues in paper drafting.
• Notes2Skills: From Lab Notebooks to Certainty-Aware Scientific Agent Skills — Preserving author certainty in informal lab notes is critical for developing reliable scientific AI agent skills.
• [D] ICML 2026 Review Discussion — ML peer review suffers from high variance in reviewer quality and increasing submission volumes.
• Improving the academic workflow: Introducing two AI agents for better figures and peer review — AI agents can automate complex academic tasks like figure generation and peer review, improving efficiency and quality.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.