Review the Code, Not the Story: A Vision and Protocol for Code-First Peer Review

2026-06-04

AI Summary

Traditional peer review relies excessively on authors’ narrative accounts, making it difficult to verify the authenticity of reported results. This work proposes a “code-first” review paradigm in which authors submit executable research artifacts alongside a checklist of claims. An AI-driven review infrastructure automatically provisions execution environments, runs experiments, audits code paths, and precisely maps each claim to empirical evidence, producing a standardized review package for human evaluation. By introducing AI as a core component of the review process, this study pioneers the concepts of claim-evidence contracts, generative review views, and the review package abstraction. It shifts the focus of peer review from narrative persuasion to verifiable, reproducible evidence and presents a comprehensive protocol framework encompassing system architecture, empirical validation, and analysis of governance challenges such as AI bias and prompt injection.

Abstract

Peer review in computational fields remains centered on author-written manuscripts, even though the decisive evidence for many claims resides in executable code, data, configurations, and experiment pipelines. This manuscript-first workflow gives authors substantial control over narrative framing while leaving reviewers with limited time to inspect implementation details, reproduce results, or detect unsupported claims. This vision and protocol paper proposes code-first peer review: authors submit executable research artifacts and minimal claim manifests; a venue-controlled AI system builds the environment, executes experiments, audits code paths, maps claims to evidence, and generates a standardized Review Package for human reviewers. The goal is not to replace reviewers or to give authors an automatic writing assistant. Instead, AI serves as review infrastructure that shifts the target of peer review from polished narratives to executable evidence. We formalize a claim-evidence contract, define the Generated Review View and Review Package abstractions, give a worked example, outline a system architecture, and analyze evaluation and governance challenges including AI bias, prompt injection, model instability, auditability, and author appeal.

Problem

• peer review
• computational research
• executable code
• reproducibility
• research integrity

Innovation

• code-first peer review
• executable research artifacts
• claim-evidence contract
• AI-assisted review infrastructure
• reproducibility