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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add knowledge graph to outline-only agent (ablation) | DualGraph (GPT-5) · DeepResearch Bench | RACE overall: 51.17 -> 53.08 | Improved |
| add knowledge graph to outline-only agent (ablation) | DualGraph (GPT-4.1) · DeepResearch Bench | RACE overall: 49.18 -> 51.50 | Improved |
| add knowledge graph guidance (ablation) | DualGraph (GPT-5) · DeepResearch Bench | FACT effective citations: 66.06 -> 79.65 | Improved |
| add knowledge graph guidance (ablation) | DualGraph (GPT-5) · DeepResearch Bench | FACT citation accuracy: 61.15 -> 57.55 | Hurt |
| add knowledge graph guidance (ablation) | DualGraph (GPT-4.1) · DeepResearch Bench II | InfoRecall: 32.18 -> 34.21 | Improved |
| add knowledge graph guidance (ablation) | DualGraph (GPT-4.1) · DeepResearch Bench II | Analysis: 50.73 -> 52.45 | Improved |
| add knowledge graph guidance (ablation) | DualGraph (GPT-4.1) · DeepConsult | win rate (%): 57.61 -> 64.42 | Improved |
| add knowledge graph guidance (ablation) | DualGraph (GPT-4.1) · DeepResearchGym | average score: 93.68 -> 94.31 | Improved |
| add knowledge graph guidance (ablation) | DualGraph (GPT-4.1) · DeepResearch Bench | mean termination iteration: 3.88 -> 3.23 | Improved |
| upgrade backbone from GPT-4.1 to GPT-5 | DualGraph · DeepResearch Bench | RACE overall: 51.50 -> 53.08 | Improved |
| add KG-enhanced outlines vs outline-only in pairwise judging (ablation) | DualGraph · DeepResearch Bench intermediate outlines | win rate (%): 53.3% at Iter 1 -> 76.2% at Iter 5 | Improved |
| remove Enrich or Explore search chains (ablation) | DualGraph · DeepResearch Bench (outline and evidence quality, Figure 3) | LLM-judge score | Hurt |
Closest papers by meaning
- WebWeaver: Structuring Web-Scale Evidence with Dynamic Outlines for Open-Ended Deep Research (ICLR 2026 · similarity 0.88)
- DAGent: Evaluate-then-Grow Planning for Deep Research Agents (NeurIPS 2026 · similarity 0.87)
- From Intent to Evidence: A Categorical Approach for Structural Evaluation of Deep Research Agents (NeurIPS 2026 · similarity 0.86)
- Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents (ICML 2026 · similarity 0.85)
- Think-on-Graph 2.0: Deep and Faithful Large Language Model Reasoning with Knowledge-guided Retrieval Augmented Generation (ICLR 2025 · similarity 0.85)
- Fathom-DeepResearch: Unlocking Long Horizon Information Retrieval and Synthesis for SLMs (ICLR 2026 · similarity 0.85)
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.
