SciAtlas: A Large-Scale Knowledge Graph for Automated Scientific Research

The exponential growth of global academic output has confronted researchers and AI agents with an unprecedented ``information explosion,'' where fragmented and unstructured knowledge organization impedes deep interdisciplinary integration. Current academic retrieval tools predominantly rely on superficial keyword matching or vector-space semantic retrieval, which lack the topological reasoning capabilities required to navigate complex logical connections. Agentic deep-research-based frameworks are often prone to logical hallucinations and consuming high inference costs. To bridge this gap, in this report, we introduce SciAtlas, a large-scale, multi-disciplinary, heterogeneous academic resource knowledge graph designed as a panoramic scientific evolution network. By integrating over 43M papers from 26 disciplines, and a total of 157M entities and 3B triplets, SciAtlas provides a structured topological cognitive substrate that dismantles disciplinary barriers and furnishes AI agents with a global perspective. Furthermore, we develop a neuro-symbolic retrieval algorithm featuring tri-path collaborative recall and graph reranking, achieving a seamless transition from simple semantic matching to deterministic association discovery. We also present key application directions of SciAtlas, including literature review, automated research trend synthesis, idea positioning, and academic trajectory exploration, to demonstrate that SciAtlas can serve as an effective ``cognitive map'' to empower the full loop of automated scientific research while significantly reducing reasoning costs. We have released the interfaces for KG retrieval and various downstream tasks in our GitHub repo.

한 줄 요약

**[Knowledge Graph / 과학 연구]** SciAtlas가 26 분과 4,300만 논문·1.57억 entity·30억 triplet의 multi-disciplinary KG 구축, neuro-symbolic tri-path recall·graph reranking으로 superficial 매칭 한계·agentic deep-research의 logical hallucination·high cost 동시 해소.

핵심 기여도

Global academic output의 exponential 성장으로 연구자·AI 에이전트가 unprecedented "information explosion"에 직면, fragmented·unstructured 지식 조직이 deep interdisciplinary 통합을 impede함을 진단.
현재 academic retrieval tool이 superficial keyword 매칭이나 vector-space semantic retrieval에 의존, 복잡 logical 연결 navigate에 필요한 topological reasoning 능력 부족; agentic deep-research 프레임워크는 logical hallucination·high inference cost에 prone.
SciAtlas 도입 — large-scale, multi-disciplinary, heterogeneous academic resource knowledge graph로 panoramic scientific evolution network로 설계. 26 분과의 4,300만+ 논문 통합, 1.57억 entity·30억 triplet 보유, 구조화 topological cognitive substrate로 분과 장벽 dismantle·AI 에이전트에 global 관점 제공.
Neuro-symbolic retrieval 알고리즘 개발 — tri-path collaborative recall·graph reranking 특징, 단순 semantic 매칭→deterministic association discovery 전환. Literature review·automated 연구 trend synthesis·idea positioning·academic trajectory exploration 등 주요 application direction 제시, SciAtlas가 "cognitive map"으로 자동 과학 연구의 full loop empower·reasoning cost 대폭 감소함을 시연. KG retrieval·various downstream task interface를 GitHub에 release.

핵심 아이디어

자동 과학 연구의 cognitive bottleneck은 단순 retrieval의 topological reasoning 부재와 agentic deep-research의 hallucination·cost에서 발생하며, multi-disciplinary 대규모 KG와 neuro-symbolic retrieval이 두 한계를 동시에 우회하는 "cognitive map" 제공한다.

기술적 접근법

**방법론**: SciAtlas — large-scale multi-disciplinary KG + neuro-symbolic retrieval.
**핵심 기법**: (1) 26 분과·4,300만+ 논문·1.57억 entity·30억 triplet의 KG 구축, (2) Topological cognitive substrate로 분과 장벽 dismantle, (3) Neuro-symbolic retrieval의 tri-path collaborative recall, (4) Graph reranking으로 deterministic association discovery, (5) Literature review·trend synthesis·idea positioning·trajectory exploration의 downstream task 인터페이스 공개.

주요 결과

26 분과·4,300만+ 논문·1.57억 entity·30억 triplet 보유 KG.
Neuro-symbolic tri-path recall·graph reranking 알고리즘.
Literature review·trend synthesis·idea positioning·trajectory exploration 효과 시연.
Agentic deep-research 대비 reasoning cost 대폭 감소.
GitHub repo 공개로 KG retrieval·downstream task 인터페이스 제공.

의의 및 한계

**의의**: 대규모 multi-disciplinary KG의 자동 과학 연구 substrate 제공, neuro-symbolic retrieval로 KG의 topological 강점·LLM의 reasoning 결합, agentic 대안 대비 cost 효율, 공개로 분야 가속. **한계**: 4,300만 논문 커버리지의 최신성·품질 유지 cost, neuro-symbolic 알고리즘의 도메인 일반화 검증, 26 분과 모두 균등 커버 여부 미상.

실용적 활용

AI co-scientist·자동 연구 시스템의 KG substrate.
학술 검색·literature review 자동화.
Interdisciplinary trend·idea discovery 도구.