AI4Research: A Survey of Artificial Intelligence for Scientific Research

TL;DR

This paper proposes AI4Research, a unified framework that classifies AI-enabled research activities into five tasks: scientific comprehension, academic survey, scientific discovery, academic writing, and academic peer reviewing. It delivers a comprehensive taxonomy of component capabilities and workflows, detailing semi-automatic and fully automatic modalities across textual, tabular, and multimodal data, as well as tool-augmented and human-in-the-loop approaches. The work catalogs methods, resources, and benchmarks, and identifies key frontiers including interdisciplinary models, ethics and safety, explainability, and multilingual/multimodal integration. By outlining abundant applications and open resources, the paper aims to accelerate AI-driven research while addressing rigor, reproducibility, and societal impact. Overall, it provides a structured roadmap for developing and evaluating AI systems that support the entire research lifecycle from discovery to dissemination.

Abstract

Recent advancements in artificial intelligence (AI), particularly in large language models (LLMs) such as OpenAI-o1 and DeepSeek-R1, have demonstrated remarkable capabilities in complex domains such as logical reasoning and experimental coding. Motivated by these advancements, numerous studies have explored the application of AI in the innovation process, particularly in the context of scientific research. These AI technologies primarily aim to develop systems that can autonomously conduct research processes across a wide range of scientific disciplines. Despite these significant strides, a comprehensive survey on AI for Research (AI4Research) remains absent, which hampers our understanding and impedes further development in this field. To address this gap, we present a comprehensive survey and offer a unified perspective on AI4Research. Specifically, the main contributions of our work are as follows: (1) Systematic taxonomy: We first introduce a systematic taxonomy to classify five mainstream tasks in AI4Research. (2) New frontiers: Then, we identify key research gaps and highlight promising future directions, focusing on the rigor and scalability of automated experiments, as well as the societal impact. (3) Abundant applications and resources: Finally, we compile a wealth of resources, including relevant multidisciplinary applications, data corpora, and tools. We hope our work will provide the research community with quick access to these resources and stimulate innovative breakthroughs in AI4Research.

Figures (9)

• Figure 1: The mainstream processes and categories of AI4Research, which can be divided into five key areas: (1) AI for Scientific Comprehension, (2) AI for Academic Survey, (3) AI for Scientific Discovery, (4) AI for Academic Writing, and (5) AI for Academic Peer Review. Each of these areas contributes to improving the effectiveness and efficiency of AI-integrated research and publication.
• Figure 2: The taxonomy of AI for research (AI4Research) is categorized into five key areas. Each area is subdivided into specific tasks, underscoring the varied roles of AI in the entire research process.
• Figure 3: The primary paradigms of AI for Scientific Comprehension. These include: (1) Textual Scientific Comprehension, which is further categorized into Semi-Automatic and Fully-Automatic Scientific Comprehension; and (2) Table & Chart Scientific Comprehension, encompassing Table and Chart Understanding.
• Figure 4: The two primary stages in AI-driven academic surveys: Related Work Retrieval and Overview Report Generation. Related Work Retrieval is further subdivided into Semantic-Guided Retrieval, Graph-Guided Retrieval, and LLM-Augmented Retrieval. Overview Report Generation encompasses Research Roadmap Mapping, Section-level Related Work Generation, and Document-level Survey Generation.
• Figure 5: The AI-augmented pipeline for scientific discovery, encompassing Idea Mining, Novelty & Significance Assessment, Theory Analysis, and Experiment Conduction. Full-Automatic Discovery integrates these elements into a cohesive, end-to-end process, supporting scientific exploration and innovation.