How can we tell when LLMs lie? We designed a domain-agnostic solution to evaluate the correctness of a Dutch LLM-based chatbot at AFAS Software, using limited data. Inspired by human decision-making, the solution is projected to save 15,000 hours per year.
Preprint arxiv.org/abs/2411.00034
In this picture, I am receiving award from Prof. Audris Mockus, and MSR 2025 core organizing team - Prof. Bram Adams and Dr. Olga Baysal
Honored to receive the MSR Rich Holt Early Career Achievement Award 2025 in Ottawa 🇨🇦, recognising my contributions impacting software companies, OSS communities, and society.
Huge thanks to my mentors, collaborators, students, friends, and family—couldn’t have done it without you! 🙏 @rug.nl
