Results of the summarizing process for the arXiv paper: 2411.13802v1

Large Language Models (LLMs) have shown promise in addressing challenges within the semiconductor industry. However, their general-purpose nature often lacks the specialized knowledge required for this sector. To fill this gap, SemiKong - the first industry-specific LLM for semiconductors - has been developed to provide a foundation for developing tailored proprietary models. With SemiKong 1.0, the focus is on understanding etching problems at an expert level. In evaluating Natural Language Generation (NLG) algorithms, human evaluation is crucial but can be costly and lack reproducibility. Automatic metrics like BLEU and ROUGE are not always reliable, leading to the introduction of LLMs as evaluators. However, these methods assume that LLMs can inherently understand and evaluate knowledge, which may not always be the case in complex domains like semiconductors. To address this limitation, a framework leveraging expert feedback is proposed to enhance assessment reliability and create a high-quality benchmark for the semiconductor domain. Semiconductor manufacturing involves intricate processes that require specialized knowledge for effective execution. Collaborating with semiconductor experts, an ontology has been developed to systematically structure semiconductor manufacturing processes. This collaboration aims to bridge the gap between AI researchers' expertise in AI and their lack of domain-specific knowledge in semiconductor manufacturing. The scope of this work includes curating a large-scale semiconductor-specific text corpus (SemiKong-Corpus), developing SemiKong as a foundation model focusing on etching problems in the semiconductor industry, fine-tuning SemiKong on industry-relevant data for process optimization and control tasks, introducing a framework to leverage expert feedback for evaluating domain-specific AI models, comparing SemiKong's performance with general-purpose LLMs, and discussing potential applications of industry-specific LLMs in semiconductor manufacturing. Overall contributions include creating a comprehensive semiconductor-related text corpus (SemiKong-Corpus), developing an industry-specific LLM (SemiKong) tailored to address specific challenges in semiconductors, advancing evaluation approaches through expert feedback integration, and highlighting the significance of industry-specific LLMs in improving AI-driven solutions for semiconductor manufacturing tasks.

• - Large Language Models (LLMs) have shown promise in addressing challenges within the semiconductor industry
• - SemiKong is the first industry-specific LLM for semiconductors, providing a foundation for developing tailored proprietary models
• - Focus of SemiKong 1.0 is on understanding etching problems at an expert level
• - Human evaluation of Natural Language Generation (NLG) algorithms is crucial but costly and lacks reproducibility
• - Automatic metrics like BLEU and ROUGE are not always reliable, leading to the introduction of LLMs as evaluators
• - Framework leveraging expert feedback proposed to enhance assessment reliability in complex domains like semiconductors
• - Collaboration with semiconductor experts to develop an ontology for structuring semiconductor manufacturing processes systematically
• - Contributions include creating a comprehensive semiconductor-related text corpus (SemiKong-Corpus), developing industry-specific LLM (SemiKong), advancing evaluation approaches through expert feedback integration, and highlighting significance of industry-specific LLMs in improving AI-driven solutions for semiconductor manufacturing tasks

Summary- Large Language Models (LLMs) are like smart computers that can help solve problems in making computer chips. - SemiKong is a special smart computer just for making computer chips, which helps create custom solutions. - SemiKong 1.0 focuses on understanding specific issues with making computer chips very well. - People need to check how good these smart computers are at writing by hand, but it's expensive and not always accurate. - Some tests used to check the smart computers aren't always right, so now they use LLMs to help. Definitions- Large Language Models (LLMs): Smart computers that understand and generate human language. - Semiconductors: Tiny electronic components used in devices like phones and computers. - Etching: A process of cutting or carving into a material, often used in semiconductor manufacturing. - Natural Language Generation (NLG): Technology that helps computers write text like humans. - Automatic metrics like BLEU and ROUGE: Tools used to measure how well a machine-generated text matches human-written text. - Ontology: A way of organizing information or knowledge in a structured manner.

Large Language Models (LLMs) have gained significant attention in recent years for their ability to process and generate large amounts of text data. These models have shown great potential in addressing challenges within various industries, including the semiconductor industry. However, due to their general-purpose nature, LLMs often lack the specialized knowledge required for this sector. To fill this gap, a team of researchers has developed SemiKong - the first industry-specific LLM for semiconductors. This model aims to provide a foundation for developing tailored proprietary models that can better understand and address complex problems specific to the semiconductor industry. The focus of SemiKong 1.0 is on understanding etching problems at an expert level. Etching is a crucial step in semiconductor manufacturing processes that involves removing layers of material from a surface using chemical reactions or physical processes. It requires specialized knowledge and expertise to optimize and control these processes effectively. One major challenge in evaluating Natural Language Generation (NLG) algorithms is the reliance on human evaluation, which can be costly and lack reproducibility. Automatic metrics like BLEU and ROUGE are not always reliable as they do not take into account domain-specific knowledge and nuances. To overcome this limitation, LLMs have been introduced as evaluators; however, they may not always possess the necessary understanding of domain-specific knowledge in complex industries like semiconductors. To address this issue, the research paper proposes a framework that leverages expert feedback to enhance assessment reliability and create a high-quality benchmark specifically for the semiconductor domain. This approach involves collaborating with semiconductor experts to develop an ontology - a structured representation of concepts within a specific domain - that systematically organizes information related to semiconductor manufacturing processes. This collaboration between AI researchers and semiconductor experts aims to bridge the gap between their respective areas of expertise - AI technology and domain-specific knowledge in semiconductor manufacturing. The scope of this work includes several key contributions: - Curating a large-scale semiconductor-specific text corpus (SemiKong-Corpus) to train industry-specific LLMs. This corpus contains a vast collection of text data related to semiconductor manufacturing processes, providing a valuable resource for developing and evaluating AI models in this domain. - Developing SemiKong as an industry-specific LLM tailored to address specific challenges in semiconductors. This model is trained on the SemiKong-Corpus and fine-tuned on industry-relevant data for process optimization and control tasks. - Introducing a framework that leverages expert feedback for evaluating domain-specific AI models. By incorporating expert knowledge, this approach aims to improve the reliability of evaluation metrics and create a high-quality benchmark specifically for the semiconductor industry. - Comparing SemiKong's performance with general-purpose LLMs to highlight the benefits of using industry-specific models in semiconductor manufacturing tasks. - Discussing potential applications of industry-specific LLMs in improving AI-driven solutions for various challenges within the semiconductor industry. Overall, this research paper makes significant contributions towards advancing AI-driven solutions in semiconductor manufacturing by creating an extensive text corpus, developing an industry-specific LLM, introducing a novel evaluation framework, and highlighting the importance of specialized knowledge in complex domains like semiconductors. In conclusion, with the development of SemiKong - the first industry-specific LLM for semiconductors - researchers have taken a step towards addressing challenges specific to this sector. By leveraging expert knowledge and developing tailored models like SemiKong, we can expect further advancements in AI technology that will greatly benefit industries such as semiconductors.