A Hierarchical Multi-Agent System for Autonomous Discovery in Geoscientific Data Archives

2026-02-24 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Environmental Science & Earth Systems · Depth: Expert, extended

Summary

PANGAEA-GPT is a hierarchical multi-agent system designed for autonomous data discovery and analysis within geoscientific data archives like PANGAEA, which hosts over 400,000 curated datasets. This framework addresses the challenge of underutilized data by implementing a Supervisor-Worker topology with data-type-aware routing, sandboxed deterministic code execution, and self-correction mechanisms. The system features a Search Agent that uses a ReAct loop for iterative query refinement, outperforming baseline keyword matching and simple LLM query translation, achieving an 8.14/10 mean score on a 100-query benchmark. Five specialist worker agents (Oceanographer, Ecologist, Visualization, DataFrame, and Writer) handle specific data types and tasks, enabling complex, multi-step workflows in physical oceanography and ecology with minimal human intervention. Validation scenarios demonstrated its capacity for cross-domain integration, statistical analysis, and visualization, including autonomously resolving API errors and refining plot layouts.

Key takeaway

For AI Researchers and Research Scientists developing autonomous systems for scientific data, PANGAEA-GPT's hierarchical multi-agent architecture offers a robust blueprint. You should consider implementing data-type-aware routing, sandboxed execution, and multi-level self-correction (programmatic and visual) to enhance system reliability and reduce manual intervention in complex, heterogeneous data environments. This approach can significantly improve data discoverability and analytical workflow automation.

Key insights

A hierarchical multi-agent system autonomously discovers and analyzes geoscientific data, improving reuse and reducing manual effort.

Principles

• Separate reasoning from execution for robust agent systems.
• Iterative query refinement enhances search precision.
• Self-correction via execution feedback improves reliability.

Method

PANGAEA-GPT employs a Supervisor-Worker architecture with data-type-aware routing to specialist agents. It uses a ReAct loop for search, sandboxed Python execution, and incorporates both programmatic traceback analysis and reflexive visual quality control for self-correction.

In practice

• Use data-type-aware routing for heterogeneous data.
• Implement sandboxed execution for code safety and state persistence.
• Integrate visual quality control for automated plot refinement.

Topics

• Hierarchical Multi-Agent Systems
• Geoscientific Data Archives
• Large Language Models
• Autonomous Data Discovery
• Self-Correction Mechanisms

Code references

• CliDyn/pangaeaGPT
• langchain-ai/langchain
• langchain-ai/langgraph

Best for: AI Researcher, AI Scientist, Research Scientist

Related on AIssential

• AutoResearchClaw: Self-Reinforcing Autonomous Research with Human-AI Collaboration — AutoResearchClaw enhances scientific discovery through iterative, multi-agent AI and strategic human collaboration.
• Prism: An Evolutionary Memory Substrate for Multi-Agent Open-Ended Discovery — Prism unifies memory paradigms under an evolutionary framework, treating memory confidence as fitness for open-ended AI discovery.
• TerraBench: Can Agents Reason Over Heterogeneous Earth-System Data? — Current LLM agents struggle with precise numerical grounding and complex workflow orchestration in Earth-science tasks.
• ANDES: Agent Native Data Evolving Synthesis Tool for Autonomous Instruction Alignment — ANDES enables AI agents to autonomously generate high-quality training data for LLM alignment by abstracting complex data curation into a steerable skill.
• Active Zero: Self-Evolving Vision-Language Models through Active Environment Exploration — Active environment exploration with co-evolving agents significantly improves VLM reasoning and general understanding.
• Article: Building Hierarchical Agentic RAG Systems: Multi-Modal Reasoning with Autonomous Error Recovery — Hierarchical multi-agent RAG with autonomous error recovery significantly improves accuracy and reduces hallucinations across structured and unstructured enterprise data.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.