Exploring a long short-term memory for mountain flood forecasting based on watershed-internal knowledge graph and large language model DOI Creative Commons
Songsong Wang, Ouguan Xu

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(3), P. e0318644 - e0318644

Published: March 13, 2025

The water levels associated with mountain floods exhibit rapid fluctuations within small watersheds, necessitating extensive data on various factors influencing such disasters to facilitate real-time forecasting. This study investigates the application of Long Short-Term Memory (LSTM) networks for flood forecasting, designing a watershed-internal Knowledge Graph (KG) and Large Language Model (LLM) that encompass watershed relationships internal information structures. We have developed hydrological KG Qixi Reservoir Qiaodongcun forecasting points located in Zhejiang Province, China, systematically organize conservancy data, identify significant disaster-related factors, optimize input determine most effective combination levels. Additionally, we implemented Recurrent Neural Networks (RNN) Gated Units (GRU) comparative analysis LSTM. findings indicate LSTM model, when integrated LLM, can effectively incorporate critical elements level changes, accuracy LLM-KG-LSTM model is enhanced by 3% compared standard series outperforms both RNN GRU models, Our method will guide future research from perspective focusing algorithms relationship between multi-dimensional disaster algorithm parallelism.

Language: Английский

Empower LlaMa 2 for Advanced Logical Reasoning in Natural Language Understanding DOI Creative Commons

Natsumi Watanabe,

Kudoro Kinasaka,

A. Nakamura

et al.

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: April 30, 2024

Abstract Enhancing LlaMa 2 through the integration of a rule-based reasoning layer and logic inference engine, collectively referred to as LlaMa+, has markedly improved its capabilities in natural language understanding. The performance enhancements were rigorously evaluated using sophisticated MMLU LegalBench benchmark datasets, where LlaMa+ showed significant increases precision, accuracy, recall, F1-score. These improvements are indicative model's advanced ability interpret reason with complex texts, reflecting substantial progression towards AI systems that can engage human contextually aware cognitively profound manner. This research not only underscores enhanced analytical abilities but also explores implications for practical applications fields requiring nuanced comprehension such legal analysis, academic research, technical problem-solving. Future directions involve optimizing computational efficiency these ensure wider applicability exploring autonomous methods updating mechanisms keep pace evolving data applications.

Language: Английский

Citations

4
Evaluating Large Language Models in Process Mining: Capabilities, Benchmarks, and Evaluation Strategies DOI
Alessandro Berti, Humam Kourani, Hannes Häfke

et al.

Lecture notes in business information processing, Journal Year: 2024, Volume and Issue: unknown, P. 13 - 21

Published: Jan. 1, 2024

Language: Английский

Citations

4
Gemini or ChatGPT? Capability, Performance, and Selection of Cutting-Edge Generative Artificial Intelligence (AI) in Business Management DOI

Nitin Rane,

Saurabh Choudhary, Jayesh Rane

et al.

SSRN Electronic Journal, Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 1, 2024

Language: Английский

Citations

4
The RL/LLM Taxonomy Tree: Reviewing Synergies Between Reinforcement Learning and Large Language Models DOI Creative Commons

Moschoula Pternea,

Prerna Singh,

Abir Chakraborty

et al.

Journal of Artificial Intelligence Research, Journal Year: 2024, Volume and Issue: 80, P. 1525 - 1573

Published: Aug. 26, 2024

In this work, we review research studies that combine Reinforcement Learning (RL) and Large Language Models (LLMs), two areas owe their momentum to the development of Deep Neural Networks (DNNs). We propose a novel taxonomy three main classes based on way model types interact with each other. The first class, RL4LLM, includes where RL is leveraged improve performance LLMs tasks related Natural Processing (NLP). RL4LLM divided into sub-categories depending whether used directly fine-tune an existing LLM or prompt LLM. second LLM4RL, assists training performs task not inherently natural language. further break down LLM4RL component framework replaces, namely reward shaping, goal generation, policy function. Finally, in third RL+LLM, agent are embedded common planning without either them contributing fine-tuning branch class distinguish between language feedback. use explore motivations behind synergy explain reasons for its success, while pinpointing potential shortcomings needed, as well alternative methodologies serve same goal.

Language: Английский

Citations

4
Exploring a long short-term memory for mountain flood forecasting based on watershed-internal knowledge graph and large language model DOI Creative Commons
Songsong Wang, Ouguan Xu

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(3), P. e0318644 - e0318644

Published: March 13, 2025

The water levels associated with mountain floods exhibit rapid fluctuations within small watersheds, necessitating extensive data on various factors influencing such disasters to facilitate real-time forecasting. This study investigates the application of Long Short-Term Memory (LSTM) networks for flood forecasting, designing a watershed-internal Knowledge Graph (KG) and Large Language Model (LLM) that encompass watershed relationships internal information structures. We have developed hydrological KG Qixi Reservoir Qiaodongcun forecasting points located in Zhejiang Province, China, systematically organize conservancy data, identify significant disaster-related factors, optimize input determine most effective combination levels. Additionally, we implemented Recurrent Neural Networks (RNN) Gated Units (GRU) comparative analysis LSTM. findings indicate LSTM model, when integrated LLM, can effectively incorporate critical elements level changes, accuracy LLM-KG-LSTM model is enhanced by 3% compared standard series outperforms both RNN GRU models, Our method will guide future research from perspective focusing algorithms relationship between multi-dimensional disaster algorithm parallelism.

Language: Английский

Citations

0