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

arXiv 2025 46.7 method

TLDR

SurveyGen-I uses coarse-to-fine retrieval, adaptive planning, and memory-guided generation to produce coherent, citation-rich scientific surveys, outperforming prior methods across four domains.

Reasoning

The paper presents a novel framework combining retrieval, adaptive planning, and memory for coherent survey generation, with empirical validation across four domains. However, it is narrowly focused on survey generation rather than broader scientific discovery, and lacks discussion of limitations or human evaluation.

Read-first score

Read-first score 46.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 56.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=experiment

Topical relevance 42%

46.7

Uses existing LLM keyword relevance scores normalized to 0-100. AI scientist,automated scientific discovery,autonomous research agent,automated research,literature review agent,survey generation,automated experimentation,experiment design agent,AI for scientific research,paper writing agent,research automation,scientific discovery agent

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 73.

Keyword Scores

survey generation

literature review agent

paper writing agent

AI for scientific research

research automation

automated research

AI scientist

automated scientific discovery

autonomous research agent

scientific discovery agent

automated experimentation

experiment design agent

Deep Analysis

Innovations

Coarse-to-fine retrieval: survey-level retrieval for initial outline and writing plan, with fine-grained subsection-level retrieval triggered when insufficient context is detected.
Adaptive planning: dynamic refinement of the outline and writing plan during generation.
Memory-guided generation: a memory mechanism that stores previously written content and terminology to maintain coherence across subsections.

Methodology

SurveyGen-I first performs survey-level retrieval to construct an initial outline and writing plan, then dynamically refines them using a memory mechanism that stores past content and terminology. When insufficient context is detected, it triggers fine-grained subsection-level retrieval. The framework is evaluated across four scientific domains.

Key Results

SurveyGen-I consistently outperforms previous works in content quality, consistency, and citation coverage across four scientific domains.

Resources

Paper