Intern-Atlas: A Methodological Evolution Graph as Research Infrastructure for AI Scientists

2026-04-30

AI Summary

This work addresses the limitation of existing research infrastructures, which are predominantly document-centric and struggle to represent the causal evolutionary relationships among research methodologies in a structured manner. To overcome this, the authors propose Intern-Atlas—the first method evolution atlas—automatically constructed from over one million AI papers. By integrating method-level entity recognition, lineage relationship inference, and evidence-grounded semantic edge construction, Intern-Atlas forms a queryable causal network comprising 9.41 million semantic edges. The authors further introduce a self-guided temporal tree search algorithm to generate methodological evolution chains. Expert evaluations confirm high consistency between the atlas-derived paths and established scientific trajectories. Intern-Atlas has already demonstrated utility in research idea evaluation and automated generation, offering a novel infrastructure for AI-powered scientific agents.

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.

Problem

• methodological evolution
• research infrastructure
• AI-driven research agents
• scientific knowledge representation
• method lineage

Innovation

• methodological evolution graph
• research infrastructure
• AI-driven scientific discovery
• temporal tree search
• method lineage