pdf material of the course "Transformer Architecture and Large Language Model Capabilities @ University of Michigan - Ann Arbor".

University of Michigan - Ann Arbor

Learners explore the foundational architecture and empirical evaluation of modern language and multimodal models. The materials cover the core mechanisms of Transformers, including multi-head attention and sequence transduction, alongside in-depth analyses of GPT-4's performance on human exams, multilingual comprehension, and visual reasoning. Participants will gain insight into predictable scaling laws, post-training alignment via RLHF and rule-based reward models, and techniques for auditing dataset contamination.

Transformer Architecture and Large Language Model Capabilities @ University of Michigan - Ann Arbor

Learn Before

Related