Compass: Navigating Global Marine Lead Data Integration through Expert-Guided LLM Agent

2026-05-28 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Robotics & Autonomous Systems · Depth: Expert, quick

Summary

Compass is an expert-guided LLM agent framework designed to integrate global marine lead (Pb) data from unstructured academic papers. Addressing the challenge of sparse in-situ observations and unscalable manual extraction, Compass utilizes a Knowledge Tree co-designed with marine scientists. This framework decomposes complex data extraction tasks into verifiable steps, guiding the LLM's reasoning to ensure scientific validity without requiring fine-tuning. Deployed across over 230,000 relevant open-access papers, Compass successfully extracted 3,751 previously unincorporated Pb records, establishing the largest integrated marine Pb database to date. The system demonstrated superior reliability with 92% accuracy, confirmed by expert manual verification. This effort expanded data coverage in under-sampled regions like the East China Sea and the Southern Ocean, providing an enriched foundation for future scientific discoveries. An interactive visualization platform has also been released.

Key takeaway

For research scientists or machine learning engineers tasked with extracting specialized data from unstructured scientific literature, consider implementing expert-guided LLM agents. This approach allows you to achieve high accuracy, demonstrated by Compass's 92% expert-verified accuracy in marine Pb data extraction, without the need for extensive fine-tuning. Your team can utilize task decomposition and domain-specific knowledge trees to ensure scientific validity and scalability, significantly accelerating data discovery in complex fields like geosciences.

Key insights

Expert-guided LLM agents can accurately extract complex scientific data from unstructured text without fine-tuning.

Principles

• LLMs require domain-specific knowledge for scientific validity.
• Task decomposition enhances LLM reasoning and verifiability.
• Expert guidance enables LLMs for high-stakes domains.

Method

The method involves an expert-guided adaptation approach for LLMs, operationalized through an agent framework with a Knowledge Tree. This decomposes tasks into verifiable steps, ensuring scientific validity without fine-tuning.

In practice

• Integrate historical data from academic papers.
• Expand data coverage in under-sampled regions.
• Create large, validated scientific databases.

Topics

• LLM Agents
• Knowledge Trees
• Marine Science
• Data Integration
• Geosciences
• Scientific Data Extraction

Best for: AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• From Data to Theory: Autonomous Large Language Model Agents for Materials Science — Autonomous LLM agents can perform end-to-end scientific theory development, from equation selection to code execution and validation.
• Meta AI - AI at Meta — Advancements span LLM reasoning, vision models, and material science, driven by new benchmarks, datasets, and training methods.
• Can Agents Price a Reaction? Evaluating LLMs on Chemical Cost Reasoning — LLM agents struggle with real-world chemical procurement cost estimation, even with tool access, due to parsing and reasoning failures.
• From Blind Guess to Informed Judgment: Teaching LLMs to Evaluate Materials by Building Knowledge-Augmented Preference Signals — Teaching LLMs to evaluate materials reliably involves pairing expert-rule-informed judgments with blind guesses as preference signals.
• Battery-Sim-Agent: Leveraging LLM-Agent for Inverse Battery Parameter Estimation — LLM agents can reframe inverse scientific problems as reasoning tasks, outperforming traditional black-box optimization.
• MIND: AI Co-Scientist for Material Research — MIND is an LLM-driven multi-agent framework for automated, experimentally verified hypothesis validation in materials science.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.