Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Atilla Kaan Alkan, Shashwat Sourav, Maja Jablonska, Simone Astarita, Rishabh Chakrabarty, Nikhil Garuda, Pranav Khetarpal, Maciej Pióro, Dimitrios Tanoglidis, Kartheik G. Iyer, Mugdha S. Polimera, Michael J. Smith, Tirthankar Ghosal, Marc Huertas-Company, Sandor Kruk, Kevin Schawinski, Ioana Ciucă

TL;DR

The paper tackles the challenge of hypothesis generation under information overload and disciplinary fragmentation, focusing on how Large Language Models (LLMs) can augment scientific discovery. It conducts a systematic literature survey spanning pre-LLM methods and LLM-driven approaches, culminating in a taxonomy that organizes human-centric, literature-based discovery, and LLM-based techniques. Key contributions include a structured overview of methodologies, evaluation strategies, and domain-specific considerations, along with a forward-looking discussion of multimodal integration and human-AI collaboration. The work provides a practical reference for researchers seeking to apply LLMs to hypothesis generation and guides future research towards robust, transparent, and ethically sound AI-assisted discovery tools.

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.

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

TL;DR

The paper tackles the challenge of hypothesis generation under information overload and disciplinary fragmentation, focusing on how Large Language Models (LLMs) can augment scientific discovery. It conducts a systematic literature survey spanning pre-LLM methods and LLM-driven approaches, culminating in a taxonomy that organizes human-centric, literature-based discovery, and LLM-based techniques. Key contributions include a structured overview of methodologies, evaluation strategies, and domain-specific considerations, along with a forward-looking discussion of multimodal integration and human-AI collaboration. The work provides a practical reference for researchers seeking to apply LLMs to hypothesis generation and guides future research towards robust, transparent, and ethically sound AI-assisted discovery tools.

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.

Paper Structure

This paper contains 26 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Paper Collection Methodology
  3. Search Strategy
  4. Inclusion Criteria
  5. Review Process
  6. Methods for Scientific Hypothesis Generation
  7. Human-Centric
  8. Literature-based Discovery
  9. Supervised Learning
  10. LLM-Driven
  11. Direct and Adversarial Prompting
  12. Fine-tuning
  13. Knowledge Integration
  14. Multi-Agent System
  15. Evaluation Methodologies
  16. ...and 11 more sections

Figures (2)

  • Figure 1: Taxonomy of Methods for Scientific Hypothesis Generation (SHG).
  • Figure 2: Pipeline of LLM-Driven Hypothesis Generation. The process begins with a research problem, which is processed by the LLM core. Various methodological branches (e.g., Direct & Adversarial Prompting, Fine-Tuning, and Knowledge Integration) contribute to a multi-agent framework that converges to generate hypotheses.