SurveyGen-I: Consistent Scientific Survey Generation with Evolving Plans and Memory-Guided Writing

2025-08-19

AI Summary

Existing LLM-driven automated literature review generation methods suffer from content inconsistency and insufficient citation coverage when generating long, multi-section scientific reviews. To address these challenges, we propose an “evolutionary planning–memory coordination” framework. It employs coarse-to-fine hierarchical retrieval to ensure both breadth and depth of source literature; integrates dynamic adaptive planning and backtracking memory mechanisms to iteratively refine writing plans and preserve cross-sectional semantic coherence during generation; and combines fine-grained retrieval triggering with context-aware generation to enhance academic rigor. Experiments across four scientific domains demonstrate that our method significantly outperforms existing baselines, achieving state-of-the-art performance on three core metrics: review coherence, logical structural completeness, and citation coverage.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Evolutionary Learning
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Models for Web evolution
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Survey papers play a critical role in scientific communication by consolidating progress across a field. Recent advances in Large Language Models (LLMs) offer a promising solution by automating key steps in the survey-generation pipeline, such as retrieval, structuring, and summarization. However, existing LLM-based approaches often struggle with maintaining coherence across long, multi-section surveys and providing comprehensive citation coverage. To address these limitations, we introduce SurveyGen-I, an automatic survey generation framework that combines coarse-to-fine retrieval, adaptive planning, and memory-guided generation. SurveyGen-I first performs survey-level retrieval to construct the initial outline and writing plan, and then dynamically refines both during generation through a memory mechanism that stores previously written content and terminology, ensuring coherence across subsections. When the system detects insufficient context, it triggers fine-grained subsection-level retrieval. During generation, SurveyGen-I leverages this memory mechanism to maintain coherence across subsections. Experiments across four scientific domains demonstrate that SurveyGen-I consistently outperforms previous works in content quality, consistency, and citation coverage.

Problem

• Maintaining coherence across long multi-section scientific surveys
• Providing comprehensive citation coverage in automated surveys
• Automating retrieval, structuring and summarization for survey generation

Innovation

• Coarse-to-fine retrieval for content coverage
• Adaptive planning with dynamic outline refinement
• Memory-guided generation ensuring cross-subsection coherence