Verified80% confidenceFactExact time
"Post-training" is the alignment and capability enhancement phase following pre-training, and typically includes SFT, RLHF, and reinforcement learning based on process reward models (such as GRPO and DAPO).
4
Sources
80%
Confidence
Long-term
Relevance
9/30/2026
First Seen
Sources
Related Entities
Related Claims
Unverified小版本号跃迁通常对应后训练（Post-Training）阶段的优化，而非预训练的重新训练78% similarVerified从预训练检查点到可商用产品需经历监督式指令微调（SFT）和基于人类反馈的强化学习（RLHF）两个后训练阶段76% similarUnverifiedA typical post-training pipeline is: pre-training → supervised fine-tuning (SFT) → preference optimization (RLHF / DPO / GRPO).75% similarUnverifiedPersona post-training typically belongs to a specialized variant of supervised fine-tuning (SFT) or preference alignment (such as RLHF, DPO), focusing more on the consistency of model outputs rather than the upper limit of capability.73% similarUnverifiedLTX-2.5 在训练中引入了更大规模的训练集与强化学习后训练（RL post-training）72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/960780API
curl https://kongchang.com/api/v1/knowledge/claims/960780MCP
get_claim(id=960780)