GRPO - Group Relative Policy Optimization: How DeepSeek Trains Reasoning Models

GRPO - Group Relative Policy Optimization: How DeepSeek Trains Reasoning Models

Serrano.Academy via YouTube Direct link

07:38 Quality Advantage

6 of 9
Previous
Next

6 of 9

07:38 Quality Advantage

Class Central Classrooms beta

YouTube videos curated by Class Central.

Classroom Contents

GRPO - Group Relative Policy Optimization: How DeepSeek Trains Reasoning Models

Automatically move to the next video in the Classroom when playback concludes
  1. 1 00:00 Introduction
  2. 2 00:26 Answering with context
  3. 3 01:40 DeepSeek vs ChatGPT
  4. 4 05:30 The GRPO score
  5. 5 07:05 Averaging over answers and steps
  6. 6 07:38 Quality Advantage
  7. 7 10:30 Probability of responses
  8. 8 15:36 Clipping the response
  9. 9 18:21 Not changing the model too much

Never Stop Learning.

Get personalized course recommendations, track subjects and courses with reminders, and more.

Sign up for free
Someone learning on their laptop while sitting on the floor.