Unverified50% confidenceFactExact time
后训练通常涵盖监督微调（SFT）、强化学习对齐（RLHF/RLAIF）以及领域适配微调，视觉推理模型还涉及视觉指令微调
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
Verified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79% similarVerifiedLarge model training goes through three core stages: pre-training, supervised fine-tuning (SFT), and preference alignment.73% similarUnverifiedThe trajectory data published by SWE-bench can be used as training signals for supervised fine-tuning (SFT) or reinforcement learning.71% similarUnverifiedModels trained with SFT (Supervised Fine-Tuning) have an 'imitation ceiling,' where their performance upper bound is constrained by the quality and diversity of the labeled data.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/530043API
curl https://kongchang.com/api/v1/knowledge/claims/530043MCP
get_claim(id=530043)