Reasoning Models in Large Language Model Inference - Lecture 9

Explore the fundamentals and advanced techniques of reasoning models in large language model inference through this comprehensive lecture from Carnegie Mellon University's Advanced NLP course. Learn what constitutes a reasoning model and discover how to train these systems using reinforcement learning approaches. Dive into the Self-Taught Reasoner (STaR) methodology and examine DeepSeek R1 alongside Group Policy Optimization (GRPO) techniques. Understand the mechanics behind long chain-of-thought reasoning processes and investigate how reasoning capabilities transfer across different domains. Master advanced reasoning algorithms including S1, L1, Stream of Search, and LAPS (Lookahead Planning with Sampling) to enhance model performance in complex problem-solving scenarios.