This paper introduces ResearchAgent, an AI system that generates research ideas by leveraging large language models (LLMs) and scientific literature. The system iteratively refines research ideas through human-aligned feedback from multiple reviewing agents. ResearchAgent begins by analyzing a core paper and its related publications, then uses an entity-centric knowledge store to extract relevant entities and concepts from scientific literature. This knowledge is used to enhance the generated research ideas. Additionally, multiple reviewing agents provide feedback based on human preferences, allowing for iterative refinement of the ideas. The system is validated across multiple disciplines, showing that it generates novel, clear, and valid research ideas that outperform existing methods. The system also demonstrates the effectiveness of using entity-centric knowledge and iterative refinement in improving the quality of research ideas. The paper also discusses limitations, including the need for further research to improve the system's ability to generate high-quality research ideas and the ethical implications of using AI in scientific research.This paper introduces ResearchAgent, an AI system that generates research ideas by leveraging large language models (LLMs) and scientific literature. The system iteratively refines research ideas through human-aligned feedback from multiple reviewing agents. ResearchAgent begins by analyzing a core paper and its related publications, then uses an entity-centric knowledge store to extract relevant entities and concepts from scientific literature. This knowledge is used to enhance the generated research ideas. Additionally, multiple reviewing agents provide feedback based on human preferences, allowing for iterative refinement of the ideas. The system is validated across multiple disciplines, showing that it generates novel, clear, and valid research ideas that outperform existing methods. The system also demonstrates the effectiveness of using entity-centric knowledge and iterative refinement in improving the quality of research ideas. The paper also discusses limitations, including the need for further research to improve the system's ability to generate high-quality research ideas and the ethical implications of using AI in scientific research.