A Survey on Hypothesis Generation for Scientific Discovery in the Era of Large Language Models

2025-04-07

AI Summary

To address low efficiency and poor interpretability in scientific hypothesis generation under information overload and disciplinary fragmentation, this study establishes the first unified methodology taxonomy for LLM-driven hypothesis generation, distinguishing two quality-enhancement pathways: novelty boosting and structured reasoning. Methodologically, it integrates prompt engineering, chain-of-thought reasoning, reflection mechanisms, retrieval-augmented generation (RAG), and multi-dimensional evaluation metrics, while proposing novel directions in multimodal fusion and interpretable human-AI collaboration. Contributions include: (1) a comprehensive knowledge graph covering methodologies, evaluation criteria, and open challenges; and (2) a theoretically grounded yet practically viable AI-augmented scientific discovery framework that significantly improves the novelty, credibility, and reusability of generated hypotheses.

Technology Category

• Search and Optimization: Metareasoning and Metaheuristics
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Generation

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Hypothesis generation is a fundamental step in scientific discovery, yet it is increasingly challenged by information overload and disciplinary fragmentation. Recent advances in Large Language Models (LLMs) have sparked growing interest in their potential to enhance and automate this process. This paper presents a comprehensive survey of hypothesis generation with LLMs by (i) reviewing existing methods, from simple prompting techniques to more complex frameworks, and proposing a taxonomy that categorizes these approaches; (ii) analyzing techniques for improving hypothesis quality, such as novelty boosting and structured reasoning; (iii) providing an overview of evaluation strategies; and (iv) discussing key challenges and future directions, including multimodal integration and human-AI collaboration. Our survey aims to serve as a reference for researchers exploring LLMs for hypothesis generation.

Problem

• Addressing information overload in scientific hypothesis generation
• Exploring LLMs' potential to automate hypothesis generation
• Improving hypothesis quality via novel techniques and frameworks

Innovation

• Surveying LLM-based hypothesis generation methods
• Improving hypothesis quality via novelty boosting
• Exploring multimodal and human-AI collaboration