Skip to content

aiXiv: A Next-Generation Open Access Ecosystem for Scientific Discovery Generated by AI Scientists

Summary

A platform paper introducing aiXiv, a next-generation open-access ecosystem designed to host research content generated by AI scientists. The authors motivate it by the mismatch between the rapid rise of LLM-driven autonomous-research systems and a publishing ecosystem that is either closed to AI-generated work (traditional journals/conferences) or quality-uncontrolled (existing preprint servers). aiXiv adopts a multi-agent architecture so that both research proposals and papers can be submitted, reviewed, and iteratively refined by both human and AI scientists, and exposes API and MCP interfaces for integrating heterogeneous human and AI participants.

Contribution

The contribution is the platform itself: an extensible multi-agent publishing ecosystem with a structured review system and iterative refinement pipelines, supporting both proposals and papers as first-class artefacts. The authors report extensive experiments demonstrating that aiXiv "significantly enhances the quality of AI-generated research proposals and papers after iterative revising and reviewing on aiXiv."

Method

System / platform paper: architecture description, comparison with existing publication platforms (arXiv, journals, conferences, Agent4Science Conference), and large-scale experimental evaluation on AI-generated proposals and papers.

Relevance to RISE

A direct ecosystem analogue of the RISE programme: aiXiv is exactly the kind of receiving venue that catalog projects like sakana-ai-scientist, research-town, zochi and reviewer systems like reviewer presuppose if their outputs are to circulate. RISE evaluation of end-to-end pipelines therefore needs to consider not just generation but the publishing substrate this paper proposes.

Critique / open questions

The platform is built and evaluated by its own authors; "significant enhancement" of generated content quality is measured against the platform's own iterative process. Reliance on AI reviewers as quality controllers risks circularity if those reviewers themselves inherit the failure modes documented in 1 and 2.

Key quotes

"This flood of AI-generated research content collides with a fragmented and largely closed publication ecosystem. Traditional journals and conferences rely on human peer review, making them difficult to scale and often reluctant to accept AI-generated research content; existing preprint servers (e.g. arXiv) lack rigorous quality-control mechanisms."

"Its multi-agent architecture allows research proposals and papers to be submitted, reviewed, and iteratively refined by both human and AI scientists."

  1. Chen, Y. et al. (2025). Reasoning models don’t always say what they think. https://arxiv.org/abs/2505.05410 ↩

  2. Matton, K. et al. (2025). Walk the talk? Measuring the faithfulness of large language model explanations. International Conference on Learning Representations (ICLR). https://arxiv.org/abs/2504.14150 ↩