SurveyX - Renmin University, in collaboration with the University of Sydney and others, launched a system for automatically generating academic reviews.
SurveyX is a system for automatically generating academic reviews based on Large Language Models (LLMs). It was jointly developed by Renmin University of China, the University of Sydney, and Northeastern University of China. Based on user-provided paper titles and keywords, it quickly generates...
What is SurveyX?
SurveyX is a system for automatically generating academic reviews based on Large Language Models (LLMs), jointly developed by Renmin University of China, the University of Sydney, and Northeastern University of China. It quickly generates high-quality, field-specific academic reviews or papers based on user-provided titles and keywords. Based on advanced language modeling technology, combined with data processing and literature retrieval capabilities, SurveyX helps users save time and effort in writing academic reviews. SurveyX breaks down the review generation process into preparation and generation stages, addressing issues such as contextual window limitations, outdated knowledge, and lack of a systematic evaluation framework in traditional methods. SurveyX outperforms existing methods in content quality, citation quality, and literature relevance, approaching the level of human experts, providing strong support for the efficient generation of high-quality academic reviews.
Main functions of SurveyX
- Automated generation of academic reviewsUsers provide the paper title and relevant keywords, and the system automatically generates high-quality academic reviews or research papers.
- Customized content generationUsers can specify the scope of literature search based on keywords according to their research needs, and generate review content in a specific field.
- Efficient Literature Retrieval and IntegrationBased on keyword retrieval of relevant literature, information from the literature is integrated to generate comprehensive and structured review content.
- Supports multiple academic fieldsThe system is applicable to a variety of disciplines, including but not limited to artificial intelligence, natural language processing, computer science, medicine, and physics.
SurveyX's technical principles
- Keyword expansion and literature retrievalBased on a keyword expansion algorithm, the search keyword pool is gradually expanded through semantic clustering and keyword extraction to ensure comprehensiveness of the search. A two-step filtering method is combined: coarse-grained filtering using an embedding model and fine-grained filtering using LLMs to ensure that the literature is highly relevant to the topic.
- Document preprocessingBy extracting key information from documents and constructing attribute trees, the information density of documents and the utilization rate of the context window of LLMs are significantly improved. Different attribute tree templates are used for different types of documents (such as methodological papers, theoretical papers, etc.) to ensure the targeting and accuracy of information extraction.
- Intelligent outline generationBased on AttributeTree, hints are generated to assist LLMs in generating secondary outlines. Redundancy is eliminated through a "separation-reorganization" process, optimizing the logical structure of the outline and ensuring the review's coherence and logical flow.
- Content generation and optimizationBased on RAG technology and combined with retrieved literature, the citation quality and accuracy of the generated content are optimized. During the generation process, LLMs can view the content of other subsections to ensure consistency of the generated content.
- Multimodal display and post-processingBased on information extraction and generation templates, necessary information is extracted from the literature to generate charts and tables, enriching the presentation of the review. Multimodal LLMs are used to retrieve supporting charts, further enhancing the readability and information delivery of the review. The generated draft is then refined to improve its fluency, logic, and academic rigor.
SurveyX project address
- Project official website:surveyx.cn
- GitHub repository:https://github.com/IAAR-Shanghai/SurveyX
- HuggingFace model library:https://huggingface.co/papers/2502.14776
- arXiv technical paper:https://arxiv.org/pdf/2502.14776
Application scenarios of SurveyX
- academic researchIt can quickly generate high-quality literature reviews, helping researchers understand the current state of the field and research directions.
- Interdisciplinary researchIntegrating multidisciplinary literature to promote the fusion of interdisciplinary knowledge.
- Dynamic updatesReal-time retrieval of the latest literature, generation of dynamic reviews, and assistance in tracking cutting-edge research.
- Teaching aidsTo help students learn review writing and improve their academic writing skills.
- Industry AnalysisGenerates technology reviews and industry reports to provide decision-making references for enterprises and institutions.