Llamarine: Open-source Maritime Industry-specific Large Language Model

TL;DR

Llamarine addresses the gap for domain-specific AI in maritime navigation by developing an open-source, maritime-focused foundation model trained through continued pretraining on a curated corpus of textbooks, regulatory texts, research papers, and Wikipedia, followed by supervised finetuning and post-training optimization. The authors introduce a dual data pipeline (pretraining and SFT) and a synthetic-plus-real evaluation benchmark to rigorously assess maritime decision-making, using navigational cues such as $SOG$, $COG$, and $CPA$ to ground evaluation. Experimental results show that Llamarine outperforms both open-source baselines and several commercial LLMs in critical maritime tasks like trajectory planning, risk assessment, and regulatory compliance, while delivering practical, context-aware guidance. The work provides an openly accessible maritime foundation model and dataset to accelerate AI-driven safety, efficiency, and operational decision-making in the industry, with plans to extend capabilities to broader maritime domains.

Abstract

Large Language Models (LLMs) have demonstrated substantial potential in addressing complex reasoning tasks, yet their general-purpose nature often limits their effectiveness in specialized domains such as maritime navigation. To bridge this gap, we introduce Llamarine, the first open-source LLM designed specifically for maritime navigation. Llamarine 1.0 is developed through continued pretraining and fine-tuning on a high-quality corpus comprising maritime textbooks, research publications, and web text from Wikipedia. This domain-specific training enables the model to acquire expert-level knowledge in navigational principles, collision avoidance, route optimization, and regulatory compliance. Our key contributions include (a) the curation of a comprehensive maritime dataset from authoritative sources, ensuring depth and reliability in the model's knowledge base; (b) the development of a foundational model capable of reasoning about complex navigational challenges with greater accuracy than general-purpose LLMs; and (c) the establishment of a benchmark to evaluate performance in maritime-specific decision-making tasks. Experimental results demonstrate that Llamarine outperforms both general-purpose and commercial LLMs in critical navigation-related tasks, such as trajectory planning, risk assessment, and compliance with maritime regulations. By providing an open-source foundation model trained exclusively on high-quality maritime literature, Llamarine paves the way for AI-driven advancements in maritime safety, efficiency, and operational decision-making.

Figures (3)

• Figure 1: Comparison of Llamarine and commercial models.. Llamarine surpasses commercial models in every aspect, including efficiency, practicality, clarity, use of examples, expert-level communication, and logicality.
• Figure 2: Question Generation Process. The system employs LLMs in a two-step process to generate maritime-domain questions. First, it synthesizes realistic scenarios from domain-specific keywords, covering navigation, vessel operations, engineering, and regulations. Optionally, additional domain knowledge can be extracted from relevant documents and sample human questions. Finally, LLMs synthesize specific questions based on these scenarios, resulting in 56,257 questions across three categories: maritime concepts (4,852), mathematical reasoning (6,065), and operational/regulatory challenges (45,340).
• Figure 3: Answer Generation Process. The process employs a two-step method where first, an LLM analyzes and breaks down the question into key components and reasoning paths. Second, these structured insights and optional related documents are used to generate the final answers.