A Comprehensive Evaluation of Multimodal Large Language Models in Hydrological Applications DOI Creative Commons

Likith Kadiyala,

Omer Mermer, R. Dinesh Jackson Samuel

и другие.

EarthArXiv (California Digital Library), Год журнала: 2024, Номер unknown

Опубликована: Май 25, 2024

Large Language Models (LLMs) combined with visual foundation models have demonstrated remarkable advancements, achieving a level of intelligence comparable to human capabilities. In this study, we conduct an analysis the latest Multimodal LLMs (MLLMs), specifically Multimodal-GPT, GPT-4 Vision, Gemini and LLaVa, focusing on their application in hydrology domain. The domain holds significant relevance for AI applications, including flood management response, water monitoring, agricultural discharge, pollution management. Our involves testing these MLLMs various hydrology-specific studies, evaluating response generation, assessing suitability real-time systems. We deliberately selected complex real-world scenarios explore potential addressing hydrological challenges. Additionally, carefully designed prompts enhance models' inference capabilities ability comprehend context from image data. findings our reveal effective human-computer interaction inspire solutions systems that incorporate both textual Among validated models, Vision stands out as top performer among other MLLMs, showcasing unparalleled proficiency inferring results highlight understanding, reasoning, decision-making multimodal bring hydrology. This research contributes valuable insights into applications advanced challenges within contexts.

Язык: Английский

End-to-End Deployment of the Educational AI Hub for Personalized Learning and Engagement: A Case Study on Environmental Science Education DOI Creative Commons
Ramteja Sajja, Yusuf Sermet, İbrahim Demir

и другие.

EarthArXiv (California Digital Library), Год журнала: 2024, Номер unknown

Опубликована: Авг. 22, 2024

This study introduces an end-to-end framework for deploying artificial intelligence (AI) enabled educational assistants tailored specifically environmental sciences learning needs in higher education. Leveraging state-of-the-art AI and natural language processing (NLP) technologies, the provides personalized experiences by facilitating access to complex data integrating seamlessly with Learning Management Systems (LMS) like Canvas Moodle. The Educational Hub agents are designed enhance course-specific utilizing innovative document parsing methods, such as Nougat technique, accurately interpret content. system offers academic support, adapting individual student extending its capabilities quantitative subjects through code execution. also emphasizes importance of accessibility, inclusivity, user privacy. results showcase potential enhanced engagement improved understanding concepts software tools, demonstrating significant impact settings, especially disciplines involving interactions. A case study, presented at 12th International Congress on Environmental Modelling Software, illustrates Hub's effectiveness improving

Язык: Английский

Процитировано

3
A Comprehensive Evaluation of Multimodal Large Language Models in Hydrological Applications DOI Creative Commons

Likith Kadiyala,

Omer Mermer, R. Dinesh Jackson Samuel

и другие.

EarthArXiv (California Digital Library), Год журнала: 2024, Номер unknown

Опубликована: Май 25, 2024

Large Language Models (LLMs) combined with visual foundation models have demonstrated remarkable advancements, achieving a level of intelligence comparable to human capabilities. In this study, we conduct an analysis the latest Multimodal LLMs (MLLMs), specifically Multimodal-GPT, GPT-4 Vision, Gemini and LLaVa, focusing on their application in hydrology domain. The domain holds significant relevance for AI applications, including flood management response, water monitoring, agricultural discharge, pollution management. Our involves testing these MLLMs various hydrology-specific studies, evaluating response generation, assessing suitability real-time systems. We deliberately selected complex real-world scenarios explore potential addressing hydrological challenges. Additionally, carefully designed prompts enhance models' inference capabilities ability comprehend context from image data. findings our reveal effective human-computer interaction inspire solutions systems that incorporate both textual Among validated models, Vision stands out as top performer among other MLLMs, showcasing unparalleled proficiency inferring results highlight understanding, reasoning, decision-making multimodal bring hydrology. This research contributes valuable insights into applications advanced challenges within contexts.

Язык: Английский

Процитировано

1