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

arXiv 2026 48.2 method

TLDR

Intern-Atlas builds a large-scale methodological evolution graph from 1M+ AI papers to help AI research agents trace method lineages.

Reasoning

Strengths include large-scale automated construction with explicit causal edges and evaluation against expert-curated ground truth. Weaknesses are domain limitation to AI and only brief mention of downstream applications.

Read-first score

Read-first score 48.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 50.

Recency 8%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=evaluation,result

Topical relevance 42%

41.7

Uses existing LLM keyword relevance scores normalized to 0-100. AI scientist,automated scientific discovery,autonomous research agent,automated research,literature review agent,survey generation,automated experimentation,experiment design agent,AI for scientific research,paper writing agent,research automation,scientific discovery agent

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 39.

Keyword Scores

AI scientist

AI for scientific research

research automation

automated scientific discovery

automated research

autonomous research agent

scientific discovery agent

literature review agent

survey generation

automated experimentation

experiment design agent

paper writing agent

Deep Analysis

Innovations

Methodological evolution graph (Intern-Atlas) that captures method-level entities, lineage relationships, and bottlenecks driving transitions between innovations
Self-guided temporal tree search algorithm for constructing evolution chains from the graph
Graph edges grounded in verbatim source evidence, forming a queryable causal network of methodological development

Methodology

Intern-Atlas is built from 1,030,314 AI papers by automatically extracting method entities, inferring lineage and bottlenecks, and creating 9.4 million semantically typed edges with verbatim evidence. A self-guided temporal tree search algorithm constructs evolution chains, which are evaluated against expert-curated ground-truth chains.

Key Results

The graph contains 9,410,201 edges; evolution chains show strong alignment with expert-curated ground truth; the infrastructure enables downstream idea evaluation and automated idea generation.

Resources

Paper