Authors: Zhen Bi, Ningyu Zhang, Yida Xue, Yixin Ou, Daxiong Ji, Guozhou Zheng, Huajun Chen

Published on: October 03, 2023

Impact Score: 8.22

Arxiv code: Arxiv:2310.02031

Summary

• What is new: Introduction of OceanGPT, the first Large Language Model (LLM) dedicated to ocean science, accompanied by a novel framework (DoInstruct) for generating instruction data and the OceanBench benchmark for evaluation.
• Why this is important: Current LLMs are inadequate for ocean science due to the complex nature of ocean data and the need for more detailed and rich knowledge.
• What the research proposes: The development of OceanGPT, a specialized LLM for ocean science, leveraging the DoInstruct framework for data generation and OceanBench for model evaluation.
• Results: OceanGPT demonstrates superior knowledge expertise in ocean science tasks and shows early signs of embodied intelligence capabilities in ocean technology.

Technical Details

Technological frameworks used: DoInstruct framework for generating instruction data, OceanBench for model evaluation

Models used: OceanGPT

Data used: A large volume of ocean domain instruction data obtained through multi-agent collaboration

Potential Impact

This research could disrupt markets related to oceanography and marine technology, benefiting companies involved in marine exploration, conservation, and the development of ocean-related technologies.

Want to implement this idea in a business?

We have generated a startup concept here: DeepSeaAI.