Research map / Multi-agent systems and games / LLM agents
WikiMAG: A Multi-Agent Guided Framework for Generating Structured Wikipedia-like Articles
Xiuli Kang, Yinlong Xiao, Minghao Hu, Yuan Huang, Bin Mao, Ming Wang, Fang Wang, Zhunchen Luo, Wei Luo, Guotong Geng
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes WikiMAG, a multi-agent framework that generates Wikipedia-like articles with structured elements (e.g., timelines, tables) beyond narrative text.
Problem
Existing LLM-based Wikipedia article generation focuses on singular narrative content and overlooks higher information-density structured elements such as timelines and tables.
Approach
A multi-agent guided framework that produces structured Wikipedia-like articles including structured elements, rather than only narrative prose.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Progressive Planner (ablation) | WikiMAG · FreshWiki | METEOR: 26.41 -> 29.27 | Improved |
| add Progressive Planner (ablation) | WikiMAG · FreshWiki | Citation Rate: 57.15 -> 61.24 | Improved |
| add Reflective Inspector (ablation) | WikiMAG · FreshWiki | METEOR: 28.33 -> 29.27 | Improved |
| add Reflective Inspector (ablation) | WikiMAG · FreshWiki | Citation Rate: 69.35 -> 61.24 | Hurt |
| add Reflective Inspector (ablation) | WikiMAG · FreshWiki | Citation Recall: 68.14 -> 68.79 | Improved |
| add Versatile Writer (ablation) | WikiMAG · FreshWiki | ROUGE-L: 20.63 -> 22.03 | Improved |
| add Versatile Writer (ablation) | WikiMAG · FreshWiki | Citation Rate: 58.48 -> 61.24 | Hurt |
| WikiMAG vs best baseline CO-STORM | WikiMAG · FreshWiki | Heading Soft Recall: 80.74 -> 87.64 | Improved |
| WikiMAG vs best baseline CO-STORM | WikiMAG · FreshWiki | Citation Recall: 51.13 -> 68.79 | Improved |
| WikiMAG vs best baseline CO-STORM | WikiMAG · WikiGenBen | METEOR: 25.62 -> 29.76 | Improved |
| reference filtering (WikiMAG vs STORM) | WikiMAG · FreshWiki | Cited/Retrieved ratio: 60% -> 72% | Improved |
Closest papers by meaning
- WikiSQE: A Large-Scale Dataset for Sentence Quality Estimation in Wikipedia (AAAI 2024 · similarity 0.85)
- Wikiformer: Pre-training with Structured Information of Wikipedia for Ad-Hoc Retrieval (AAAI 2024 · similarity 0.82)
- WebWeaver: Structuring Web-Scale Evidence with Dynamic Outlines for Open-Ended Deep Research (ICLR 2026 · similarity 0.82)
- EMERGE: A Benchmark for Updating Knowledge Graphs with Emerging Textual Knowledge (NeurIPS 2026 · similarity 0.82)
- Learning Pattern-Based Extractors from Natural Language and Knowledge Graphs: Applying Large Language Models to Wikipedia and Linked Open Data (AAAI 2024 · similarity 0.82)
- AutoData: A Multi-Agent System for Open Web Data Collection (NeurIPS 2025 · similarity 0.81)
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.
