Paul Tarau on a minimal logic for constructive next-token inference
Автор: Logic Programming Education
Загружено: 2026-03-03
Просмотров: 10
Описание:
Paul Tarau takes on the research quest to describe the foundations of today's GenAI with the simplest possible formalism -- so that inference steps constructively model the next-token prediction LLMs are based on.
This project includes some experiments on training, learning and reasoning with LLM inspired logical systems, to show the correspondence.
Paul Tarau is an active collaborator of PEG 2.0 and very well respected member of the logic programming community. His commitment to explore connections with other technologies is inspiring and he has shown that there are roads to the creation of better systems that combine them.
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
