Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Graph Your Way to Inspiration: Integrating Co-Author Graphs with Retrieval-Augmented Generation for Large Language Model Based Scientific Idea Generation

Pengzhen Xie, Huizhi Liang

TL;DR

A scientific idea generation system called GYWI, which combines author knowledge graphs with retrieval-augmented generation (RAG) to form an external knowledge base to provide controllable context and trace of inspiration path for LLMs to generate new scientific ideas is proposed.

Abstract

Large Language Models (LLMs) demonstrate potential in the field of scientific idea generation. However, the generated results often lack controllable academic context and traceable inspiration pathways. To bridge this gap, this paper proposes a scientific idea generation system called GYWI, which combines author knowledge graphs with retrieval-augmented generation (RAG) to form an external knowledge base to provide controllable context and trace of inspiration path for LLMs to generate new scientific ideas. We first propose an author-centered knowledge graph construction method and inspiration source sampling algorithms to construct external knowledge base. Then, we propose a hybrid retrieval mechanism that is composed of both RAG and GraphRAG to retrieve content with both depth and breadth knowledge. It forms a hybrid context. Thirdly, we propose a Prompt optimization strategy incorporating reinforcement learning principles to automatically guide LLMs optimizing the results based on the hybrid context. To evaluate the proposed approaches, we constructed an evaluation dataset based on arXiv (2018-2023). This paper also develops a comprehensive evaluation method including empirical automatic assessment in multiple-choice question task, LLM-based scoring, human evaluation, and semantic space visualization analysis. The generated ideas are evaluated from the following five dimensions: novelty, feasibility, clarity, relevance, and significance. We conducted experiments on different LLMs including GPT-4o, DeepSeek-V3, Qwen3-8B, and Gemini 2.5. Experimental results show that GYWI significantly outperforms mainstream LLMs in multiple metrics such as novelty, reliability, and relevance.

Graph Your Way to Inspiration: Integrating Co-Author Graphs with Retrieval-Augmented Generation for Large Language Model Based Scientific Idea Generation

TL;DR

A scientific idea generation system called GYWI, which combines author knowledge graphs with retrieval-augmented generation (RAG) to form an external knowledge base to provide controllable context and trace of inspiration path for LLMs to generate new scientific ideas is proposed.

Abstract

Large Language Models (LLMs) demonstrate potential in the field of scientific idea generation. However, the generated results often lack controllable academic context and traceable inspiration pathways. To bridge this gap, this paper proposes a scientific idea generation system called GYWI, which combines author knowledge graphs with retrieval-augmented generation (RAG) to form an external knowledge base to provide controllable context and trace of inspiration path for LLMs to generate new scientific ideas. We first propose an author-centered knowledge graph construction method and inspiration source sampling algorithms to construct external knowledge base. Then, we propose a hybrid retrieval mechanism that is composed of both RAG and GraphRAG to retrieve content with both depth and breadth knowledge. It forms a hybrid context. Thirdly, we propose a Prompt optimization strategy incorporating reinforcement learning principles to automatically guide LLMs optimizing the results based on the hybrid context. To evaluate the proposed approaches, we constructed an evaluation dataset based on arXiv (2018-2023). This paper also develops a comprehensive evaluation method including empirical automatic assessment in multiple-choice question task, LLM-based scoring, human evaluation, and semantic space visualization analysis. The generated ideas are evaluated from the following five dimensions: novelty, feasibility, clarity, relevance, and significance. We conducted experiments on different LLMs including GPT-4o, DeepSeek-V3, Qwen3-8B, and Gemini 2.5. Experimental results show that GYWI significantly outperforms mainstream LLMs in multiple metrics such as novelty, reliability, and relevance.
Paper Structure (56 sections, 4 equations, 9 figures, 1 table, 2 algorithms)

This paper contains 56 sections, 4 equations, 9 figures, 1 table, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. The proposed Approach: GYWI for Idea Generation
  4. author-centered Knowledge Graph
  5. Knowledge Graph Construction
  6. Inspiration Source Sampling Algorithm
  7. Parameterisation
  8. Hybrid Retrieval for Depth and Breadth
  9. Adaptive Prompt Optimization for Idea Generation
  10. Prompt Optimization Strategy
  11. Experiments
  12. Evaluation Method
  13. Dataset construction
  14. Multi-dimensional Scoring Assessment
  15. Experimental Results and Analysis
  16. ...and 41 more sections

Figures (9)

  • Figure 1: Author-guided scientific research idea generation system Note: This system combines author knowledge graphs with hybrid retrieval methods to generate scientific research ideas.
  • Figure 2: Author Knowledge Graph (generated example). This auto-generated graph illustrates the collaboration network derived from co-authorships around a target paper.
  • Figure 3: Dataset Statistics. (a) Abstracts are mostly between 120–180 words, providing sufficient context for motivation extraction. (b) IMCQ options are length-balanced, avoiding answer bias.
  • Figure 4: Accuracy of GYWI based on different LLMs in the IMCQ evaluation.
  • Figure 5: IMCQ Accuracy Improvement by Module.
  • ...and 4 more figures

Theorems & Definitions (3)

  • Example 1: Author and Paper Sets
  • Example 2: Adjacent Sampling Around Target Paper
  • Example 3: Minimal illustrations of the hybrid retrieval