AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

A Tale of Two Graphs: Separating Knowledge Exploration from Outline Structure for Open-Ended Deep Research

Zhuofan Shi, Ming Ma, Zekun Yao, Fangkai Yang, Jue Zhang, Dongge Han, Victor Rühle, Qingwei Lin, Saravan Rajmohan, Dongmei Zhang

ICML 2026 · Graph learning › Knowledge and temporal graphs · approach: Temporal knowledge graph reasoning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add knowledge graph to outline-only agent (ablation)DualGraph (GPT-5) · DeepResearch BenchRACE overall: 51.17 -> 53.08Improved
add knowledge graph to outline-only agent (ablation)DualGraph (GPT-4.1) · DeepResearch BenchRACE overall: 49.18 -> 51.50Improved
add knowledge graph guidance (ablation)DualGraph (GPT-5) · DeepResearch BenchFACT effective citations: 66.06 -> 79.65Improved
add knowledge graph guidance (ablation)DualGraph (GPT-5) · DeepResearch BenchFACT citation accuracy: 61.15 -> 57.55Hurt
add knowledge graph guidance (ablation)DualGraph (GPT-4.1) · DeepResearch Bench IIInfoRecall: 32.18 -> 34.21Improved
add knowledge graph guidance (ablation)DualGraph (GPT-4.1) · DeepResearch Bench IIAnalysis: 50.73 -> 52.45Improved
add knowledge graph guidance (ablation)DualGraph (GPT-4.1) · DeepConsultwin rate (%): 57.61 -> 64.42Improved
add knowledge graph guidance (ablation)DualGraph (GPT-4.1) · DeepResearchGymaverage score: 93.68 -> 94.31Improved
add knowledge graph guidance (ablation)DualGraph (GPT-4.1) · DeepResearch Benchmean termination iteration: 3.88 -> 3.23Improved
upgrade backbone from GPT-4.1 to GPT-5DualGraph · DeepResearch BenchRACE overall: 51.50 -> 53.08Improved
add KG-enhanced outlines vs outline-only in pairwise judging (ablation)DualGraph · DeepResearch Bench intermediate outlineswin rate (%): 53.3% at Iter 1 -> 76.2% at Iter 5Improved
remove Enrich or Explore search chains (ablation)DualGraph · DeepResearch Bench (outline and evidence quality, Figure 3)LLM-judge scoreHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy