-
Analyzing constrained LLM through PDFA-learning
Abstract: We define a congruence that copes with null next-symbol probabilities that arise when the output of a language model is constrained by some means during text generation. We develop an algorithm for efficiently learning the quotient with respect to this congruence and evaluate it on case studies for analyzing statistical properties of LLM.
Submitted 15 June, 2024; v1 submitted 12 June, 2024; originally announced June 2024.
Comments: Workshop Paper
-
Embracing Agile methodology during DevOps Developer Internship Program
Abstract: The DevOps team adopted agile methodologies during the summer internship program as an initiative to move away from waterfall. The DevOps team implemented the Scrum software development strategy to create an internal data dictionary web application. This article reports on the transition process and lessons learned from the pilot program.
Submitted 7 July, 2016; originally announced July 2016.
Comments: 5 pages, 1 figure, Submitted for peer review to IEEE Software