Unverified50% confidenceFactExact time
GRPO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
1
Sources
50%
Confidence
Long-term
Relevance
7/22/2026
First Seen
Sources
Related Claims
UnverifiedGRPO 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.78% similarUnverifiedDeepSeek explored the GRPO algorithm as an attempt to improve upon the original RLHF framework.75% similarUnverifiedQwen2.5后训练首次在在线阶段使用GRPO，GRPO是DeepSeek团队提出的PPO轻量化变体，去掉独立价值网络，用组内平均奖励作为基线74% similarUnverifiedDeepSeek-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74% similarVerifiedDeepSeek-R1通过在代码和数学两个可验证域上大规模运行GRPO算法取得成功73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/586990API
curl https://kongchang.com/api/v1/knowledge/claims/586990MCP
get_claim(id=586990)