Post-Training Techniques: How LLMs Learn to Follow Instructions
CME295 Transformers and Large Language Models is open for enrollment until September 6. Learn more: https://online.stanford.edu/courses/cme295-transformers-and-large-language-models
A pretrained LLM is good at predicting text — but that's not the same as following instructions well. Stanford Adjunct Professors Shervine and Afshine Amidi walk through the techniques that close that gap: fine-tuning, reinforcement learning, preference optimization, verifier-guided training, and distillation.
CME296 Diffusion and Large Vision Models will be available in Spring 2027. Learn more about that course here: https://online.stanford.edu/courses/cme296-diffusion-and-large-vision-models
#PostTraining #LLM #Transformers #AI #MachineLearning #ReinforcementLearning #OnlineCourses
A pretrained LLM is good at predicting text — but that's not the same as following instructions well. Stanford Adjunct Professors Shervine and Afshine Amidi walk through the techniques that close that gap: fine-tuning, reinforcement learning, preference optimization, verifier-guided training, and distillation.
CME296 Diffusion and Large Vision Models will be available in Spring 2027. Learn more about that course here: https://online.stanford.edu/courses/cme296-diffusion-and-large-vision-models
#PostTraining #LLM #Transformers #AI #MachineLearning #ReinforcementLearning #OnlineCourses
Stanford Online
You can gain access to a world of education through Stanford Online, the Stanford School of Engineering’s portal for academic and professional education offered by schools and units throughout Stanford University. https://online.stanford.edu/
Our robust ...