GRPO doesn't remove the reward model. It removes the critic.
This story is from 2026-09-18. It is preserved in the archive; the latest stories are on the live feed.
Every time GRPO comes up I see the same slip — someone says it "gets rid of the reward model". It doesn't. It gets rid of the value model. Those are two different networks doing unrelated jobs, and telling them apart is most of understanding the algorithm. Worth going back to what the DeepSeekMath…
Read the full story at DEV Community — Machine Learning ↗
Timeline · 1 report
- 2026-09-18 04:00 · DEV Community — Machine Learning
GRPO doesn't remove the reward model. It removes the critic.