Intern-Atlas: A Methodological Evolution Graph as Research Infrastructure for AI Scientists
Yujun Wu, Dongxu Zhang, Xinchen Li, Jinhang Xu, Yiling Duan, Yumou Liu, Jiabao Pan, Xuanhe Zhou, Jingxuan Wei, Siyuan Li, Jintao Chen, Conghui He, Cheng Tan
2026-05-01
Summary
This paper introduces a new way to map how research methods in Artificial Intelligence have changed and developed over time, creating a detailed 'family tree' of ideas.
What's the problem?
Currently, most research databases focus on which papers cite others, but they don't show *how* methods within those papers actually evolve. They lack a clear understanding of how one technique leads to another, or why certain approaches become popular while others fade away. This is becoming a big issue with the rise of AI tools that try to understand and build upon existing research, because those tools need to know the relationships between methods, not just which papers mention them.
What's the solution?
The researchers created 'Intern-Atlas,' a massive graph that connects over 9 million relationships between different AI methods, based on analyzing more than a million research papers. This graph doesn't just say method A influenced method B; it also provides specific evidence from the papers themselves to support that connection. They also developed a way to automatically trace the history of a method, showing how it changed step-by-step over time.
Why it matters?
This work is important because it provides a structured foundation for AI systems to automatically explore and understand the history of research. It allows these systems to not only find ideas but also evaluate them and even generate new ones, ultimately speeding up the process of scientific discovery. It's like giving AI the ability to learn from the past, not just read about it.
Abstract
Existing research infrastructure is fundamentally document-centric, providing citation links between papers but lacking explicit representations of methodological evolution. In particular, it does not capture the structured relationships that explain how and why research methods emerge, adapt, and build upon one another. With the rise of AI-driven research agents as a new class of consumers of scientific knowledge, this limitation becomes increasingly consequential, as such agents cannot reliably reconstruct method evolution topologies from unstructured text. We introduce Intern-Atlas, a methodological evolution graph that automatically identifies method-level entities, infers lineage relationships among methodologies, and captures the bottlenecks that drive transitions between successive innovations. Built from 1,030,314 papers spanning AI conferences, journals, and arXiv preprints, the resulting graph comprises 9,410,201 semantically typed edges, each grounded in verbatim source evidence, forming a queryable causal network of methodological development. To operationalize this structure, we further propose a self-guided temporal tree search algorithm for constructing evolution chains that trace the progression of methods over time. We evaluate the quality of the resulting graph against expert-curated ground-truth evolution chains and observe strong alignment. In addition, we demonstrate that Intern-Atlas enables downstream applications in idea evaluation and automated idea generation. We position methodological evolution graphs as a foundational data layer for the emerging automated scientific discovery.