About Us
We are Robert and Kevin, two Postdoctoral Researchers at ETH Zurich and Directors of the Agentic Systems Lab. During our doctoral studies, we became deeply frustrated by how the formal peer review process at conferences and journals not only slows down the dissemination of scientific knowledge, but also diverts valuable time and resources away from actual scientific discovery.
As a starting point, we built Rigorous v0.1, a multi-agent architecture with dozens of specialized AI reviewer agents. We first tested the system internally during one of our ETH chair's legendary Doctoral Seminars in April 2025. A few weeks later, we launched it publicly on Hacker News, where it unexpectedly went viral resulting in hundreds of manuscript submissions that pushed our system to its limits. At the same time, it sparked interesting discussions about the structural issues in academia and the future of scientific publishing.
Since then, we’ve been approached by PhD students, postdocs, professors, publishing houses, investors, and others across the academic ecosystem. Through these conversations, we've come to better understand the deep structural challenges facing science—challenges that we expect to grow more acute in the years ahead due to increased publication volume, incentives misalignment, and AI-generated content.
With the Rigorous project, we're committed to building AI-native workflows that accelerate, improve, and democratize the scientific process.
🧩 Problem
The current system of academic peer review—long regarded as the gold standard for scientific validation—is under growing strain. An exponential rise in global research output, coupled with stagnant peer review infrastructure, has created a system that is:
- Slow: Manuscripts often wait months or even years for evaluation and publication.
- Expensive: High publication fees lock out underfunded researchers and divert resources away from actual scientific discovery.
- Opaque and inconsistent: Peer reviews vary widely in quality and transparency, with reviewer fatigue and bias undermining trust.
- Vulnerable: The rise of AI-assisted "papermills" and predatory journals threatens the credibility of published science.
- Misaligned: Researchers are frequently solicited via impersonal spam emails to contribute reviews, yet rarely receive recognition, incentives, or compensation for their contribution—despite its central role in maintaining scientific quality.
At the same time, the complexity and volume of modern research outpaces human capacity for thorough, timely, and scalable review.
🔍 Research Aim
We investigate the use of large language models (LLMs) and agentic AI systems as co-reviewers and assistive agents in the scientific peer review process. The goal is to assess whether and how AI can:
- Accelerate scientific dissemination by automating routine evaluation tasks
- Enhance transparency and consistency through structured reasoning and auditability
- Reduce the cost of scholarly publishing via effective, reproducible review pipelines
- Improve early-stage manuscript feedback to support research development worldwide
- Support the paper/reviewer matchmaking process
🔬 Research Approach
We explore and experimentally evaluate:
- AI review agents that assess methodology, clarity, and reproducibility of scientific manuscripts
- Multi-agent architectures that mimic diverse reviewer perspectives across disciplines
- Explainable and auditable AI systems to ensure trust, traceability, and alignment with scholarly norms
- Benchmarking frameworks to evaluate AI reviewer performance
- Human-AI collaboration models to balance automated screening with expert judgment
- Economic models to study cost efficiency, particularly in resource-constrained settings
🌱 Anticipated Outcomes and Impact
The project aims to contribute foundational insights into how AI can responsibly augment—not replace—the human peer review process. Expected outcomes include:
- Empirical evidence on the strengths and limitations of LLMs and Agentic Systems in manuscript evaluation
- Tools and protocols for integrating AI into early-stage review and journal workflows
- Frameworks for transparency and governance to address ethical, security, and bias concerns
- New pathways for identifying overlooked errors and gaps in existing scientific literature through AI review
Long-term, this research aspires to support a more inclusive, transparent, and efficient global research ecosystem, where the barriers to contributing and accessing rigorous science are dramatically lowered.
🧠 Why This Matters to Us
Scientific progress depends on rigorous evaluation and timely dissemination of knowledge. By reimagining peer review as a human–AI collaborative process, we seek to:
- 🛠 Enable more researchers to share high-quality science faster
- 🔍 Preserve and enhance the integrity of scientific communication
- 🌍 Democratize access to high-quality feedback
- 🧯 Reveal overlooked errors and limitations in the existing body of scientific knowledge
Most importantly, this project is not about automating judgment—but about scaling support, preserving rigor, and reclaiming researchers' time for creativity, discovery, and deep review.
Meet the Team
Disclaimer: This project is currently run by us in our personal capacity and thus independent from our ETH research roles.