Скачать или смотреть 4 Examination of the Reliability of Generated LLM code, Prof. Foutse Khomh

Examination of the Reliability of Generated LLM Code

About the Speaker
Professor Foutse Khomh, a full professor at Polytechnique Montreal, is renowned for his contributions to software engineering. He holds the NSERC author B McDonald Fellowship and the Canadian CIFAR AI chair, among other distinguished titles. His impactful work has earned him numerous awards, including the Outstanding Early Career Computer Science Researcher Award from CSCAN/INFO-CAN in 2019.

About the Talk
In his keynote in the FM Generated Code and FMware session, Professor Khomh delved into the challenges and potential of using large language models (LLMs) for code generation. He presented ongoing research focused on enhancing the reliability of code produced by these models. Highlighting the rapid development and increasing adoption of LLMs in software engineering, he pointed out that despite their advanced capabilities, there are significant concerns regarding the reliability of the generated code. Through his research, Professor Khomh has been analyzing common fault patterns in code generated by LLMs, aiming to develop methodologies that improve their reliability. His findings suggest that these models often fail in ways that are fundamentally different from human programmers, indicating a need for tailored quality assurance approaches to better manage and utilize LLM-generated code in practice.

https://fmse.io/