AI & Machine Learning
Post Training Strategy
为 Agent 模型做后训练或选路线时使用——判断该先补 Mid-training、用 SFT 立协议还是直接上 RL；构造 Mid-training 语料与 SFT 数据、搭建 RL 环境、在 GRPO/PPO/DPO 与蒸馏之间取舍；排查 pass@k 近零、格式不稳就训 RL、组内无奖励差异、训练-推理数值失配、过度训练泛化下降。触发词：后训练、Mid-training、继续预训练、SFT、RL、GRPO、PPO、DPO、拒绝采样、RFT、蒸馏、on-policy、RLVR、LoRA。
- Publisher / source
- bojieli/ai-agent-book (found on GitHub top repositories)
- Upstream record
- bojieli/ai-agent-book
- Version
- Version 1
- Published
- Publication review
- Approved for publication · recorded passed scan 2026-11-03
- Content fingerprint
34b7b049bb42a414ff0f0cd8717147561d70c67c1aa3d787cf82dd807aef13c4