aiXiv launches as next-generation open-access platform for AI scientists — research proposals submitted, reviewed, and refined by both humans and AI agents
aiXiv was introduced as a next-generation open-access ecosystem for scientific discovery generated by AI scientists. The multi-agent architecture allows research proposals and papers to be submitted, reviewed, and iteratively refined by both human and AI scientists. It's the first major open-access platform that treats AI scientists as first-class submitters and reviewers rather than as tools used by human authors.
The first-class-AI-scientist framing is the structural shift. Through 2024-2025 the ML research community split roughly into two camps on AI-author papers — one that accepted AI-coauthored papers under disclosure rules (most journals and arXiv categories), and one that treated AI-generated submissions with suspicion or outright rejection. aiXiv's position is that the AI scientist is a research participant whose submissions stand or fall on the same review process as a human's. Whether the field accepts that framing will become visible through 2026 in the citation rates and adoption patterns of aiXiv-originated work.
The multi-agent review architecture is the implementation that makes the framing tractable. Submissions enter a pipeline where multiple specialized reviewer agents critique the work — methodology, claims-versus-evidence, related-literature coverage, reproducibility — and human reviewers sit on top of the agent outputs. The system is designed to be auditable: each agent's review is reproducible from the agent's prompt and the paper text, and the human reviewer's role is to evaluate the agent reviews as much as the original paper. It's a more transparent version of peer review than the human-only black box that journals have used for centuries.
ArXiv — aiXiv Next-Generation Open Access Ecosystem → · ArXiv — Artificial Intelligence Recent Submissions → · ArXiv — Mechanistic Interpretability for AI Safety Review →