A naukri.com initiative
Medium
1M
40
Image Credit: Medium
What makes DeepSeek Math solve IMO questions
- DeepSeek Math uses a rule-based reward system to solve IMO questions.
- The reward system includes accuracy rewards and format rewards.
- DeepSeek Math adopted Proximal Policy Optimization (PPO) and Group Relative Policy Optimization (GRPO).
- The model prevents catastrophic forgetting of the learned LLMs by bounding policy updates.
Read Full Article
2 Likes
For uninterrupted reading, download the app