Unverified50% confidenceFactExact time
DeepSeek-R1等工作探索了以GRPO替代PPO，通过组内相对评分简化价值模型估计，降低四模型并行的显存压力
1
Sources
50%
Confidence
Long-term
Relevance
7/18/2026
First Seen
Sources
Related Claims
UnverifiedDeepSeek explored the GRPO algorithm as an attempt to improve upon the original RLHF framework.75% similarVerifiedGRPO由DeepSeek团队系统化提出并在训练DeepSeek-R1系列模型时得到大规模验证74% similarUnverifiedGRPO和GSPO均是PPO算法的改进变体，DeepSeek-R1凭借GRPO在推理能力上实现大幅跃升74% similarUnverifiedGRPO is a reinforcement learning algorithm proposed by the DeepSeek team that eliminates the Critic Model from PPO, computing intra-group relative rewards as a baseline by generating a set of responses to the same question.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/550663API
curl https://kongchang.com/api/v1/knowledge/claims/550663MCP
get_claim(id=550663)