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

ReasFlow: Assisting Reasoning-Centric Scientific Discovery in Applied Mathematics via a Knowledge-Based Multi-Agent System

arXiv 2026 78.9 method

TLDR

ReasFlow is a multi-agent system for reasoning-centric scientific discovery in applied mathematics, automating proofs, synthesis, and paper generation.

Reasoning

Strengths: Addresses a gap in theory-driven discovery with a robust verification loop and knowledge retrieval. Weaknesses: Limited to applied mathematics; generalizability uncertain. The abstract is cut off, but the system appears to generate full papers, indicating real-world testing.

Read-first score

Read-first score 78.9, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 93.

Recency 8%
100

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

Methodology quality 25%
80

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

Topical relevance 42%
77.5

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%
73

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

Field roles

FrontierMethodology anchorReproducibility anchor

Rank sensitivity

Stability: sensitive; rank range: 3.

Keyword Scores

automated scientific discovery
9
autonomous research agent
9
AI for scientific research
9
research automation
9
scientific discovery agent
9
AI scientist
8
automated research
8
paper writing agent
8
automated experimentation
7
literature review agent
6
experiment design agent
6
survey generation
5

Deep Analysis

Innovations

  • End-to-end autonomous agent system for reasoning-centric scientific discovery in applied mathematics, targeting theory-driven domains underserved by existing empirical research agents.
  • Internal verification loop that audits logical coherence and corrects fundamental errors prior to human inspection.
  • Automated knowledge retrieval and self-improvement mechanism that surfaces both declarative facts and overlooked procedural heuristics to reduce expert intervention.
  • Unified pipeline integrating literature synthesis, algorithm design, theorem proving, experimentation, and manuscript preparation.

Methodology

ReasFlow is a multi-agent system that frames human–AI collaboration as Principal Investigator (human) and graduate student (agent). It uses an internal verification loop for logical error correction and a knowledge retrieval module for self-improvement. The system was evaluated by autonomously generating five complete research papers from minimal prompts, then scoring them against open-access baselines with a curated LLM-based review rubric.

Key Results

ReasFlow consistently achieved the highest evaluation scores among state-of-the-art open-access baselines across five generated papers containing rigorous theoretical and empirical content.

Tags