Awesome Auto Research Hub Papers · Datasets · Projects
← Back to papers

ResearchEVO: An End-to-End Framework for Automated Scientific Discovery and Documentation

arXiv 2026 64.4 method

TLDR

ResearchEVO is an end-to-end framework that combines LLM-guided algorithm evolution with autonomous paper generation for automated scientific discovery.

Reasoning

The paper presents a novel two-phase framework (evolution and writing) validated on real-world problems, demonstrating strong empirical results. However, the abstract lacks details on limitations and comparison to baselines, and the writing phase's anti-hallucination verification is not fully evaluated.

Read-first score

Read-first score 64.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 98.

Recency 8%
100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Topical relevance 42%
81.7

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

Methodology quality 25%
50

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

Reproducibility 25%
38

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 135.

Keyword Scores

automated scientific discovery
10
AI scientist
9
automated research
9
automated experimentation
9
AI for scientific research
9
paper writing agent
9
research automation
9
scientific discovery agent
9
autonomous research agent
8
experiment design agent
8
literature review agent
5
survey generation
4

Deep Analysis

Innovations

  • End-to-end framework automating the full discover-then-explain scientific paradigm, from blind algorithm evolution to publication-ready paper generation.
  • LLM-guided bi-dimensional co-evolution that simultaneously optimizes algorithmic logic and overall architecture purely by fitness, without requiring prior understanding.
  • Sentence-level retrieval-augmented generation with explicit anti-hallucination verification and automated experiment design for writing phase.
  • First system to jointly perform principled algorithm evolution and literature-grounded scientific documentation in a single pipeline.

Methodology

ResearchEVO consists of an Evolution Phase that uses LLM-guided bi-dimensional co-evolution to search code implementations by fitness, and a Writing Phase that generates a complete LaTeX manuscript via sentence-level retrieval-augmented generation with anti-hallucination checks and automated experiment design. The framework is validated on Quantum Error Correction using real Google quantum hardware data and on Physics-Informed Neural Networks.

Key Results

The Evolution Phase discovered human-interpretable algorithmic mechanisms not previously proposed in the domain literatures for both tasks. The Writing Phase autonomously produced compilable LaTeX manuscripts that correctly grounded the blind discoveries in existing theory via RAG, with zero fabricated citations.

Tags

AIOC