Skip to main content
Articles tagged with deepseek | Vitor Sousa
Vitor Sousa
Writing
Projects
Foundations
Start Here
Bits
⌘K
Writing
Projects
Foundations
Start Here
Bits
Toggle theme
Suggested searches
transformers
evaluation
reinforcement learning
contextual bandits
Articles tagged deepseek
Back to blog
→
Feb 3
GRPO: Eliminating the Value Network
Part 3
Group Relative Policy Optimization replaces PPO's learned value function with a simple insight: sample multiple outputs and use their relative rewards as advantages. 33% memory savings, simpler implementation, and the algorithm powering DeepSeek-R1.
~32 min