Expired50% confidenceFactExact time
GRPO emerged in 2024 and was adopted by DeepSeek-R1, further simplifying the reference model dependency on top of DPO
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/6/2026
First Seen
Valid until: 10/4/2026(expired)
Sources
Programmers Pivoting to AI: Opportunities in Application Development and Pitfalls to Avoid
bilibili码士集团-Java技术库6/5/2026
Related Claims
UnverifiedGRPO was proposed by the DeepSeek team in 2024 and is a lightweight variant of PPO.81% similarUnverifiedGrok-1 was released in open-source form in March 2024, with 314 billion parameters, using a Mixture of Experts (MoE) architecture.68% similarUnverifiedSince 2024, further simplified alignment algorithms such as ORPO, SimPO, and KTO have emerged.63% similarUnverified2024年底至2025年初，DeepSeek-V3和DeepSeek-R1相继发布62% similarUnverifiedThe combination of LoRA + DPO + ReAct had, by 2024, almost become the standard recipe for Agent projects.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/112698API
curl https://kongchang.com/api/v1/knowledge/claims/112698MCP
get_claim(id=112698)