Unverified50% confidenceFactExact time
LongZero uses the GRPO algorithm to train the model, mixing the two task types of multi-hop reasoning and multi-turn retrieval.
1
Sources
50%
Confidence
Long-term
Relevance
10/22/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedGRPO (Group Relative Policy Optimization), proposed by DeepSeek, requires no separate training of a value network and has become the mainstream algorithm choice for current agentic RL training.75% similarUnverifiedDeepSeek extensively uses the GRPO algorithm in place of traditional RLHF during the training phase to reduce training costs.72% 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72% similarUnverifiedDeepSeek团队报告，仅用GRPO在数学和代码数据上进行强化学习而不经过监督微调，模型便自发出现反思、回溯验证等链式推理行为（aha moment）71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1084158API
curl https://kongchang.com/api/v1/knowledge/claims/1084158MCP
get_claim(id=1084158)