微调 · 换上下文和改模型,不是一回事Fine-tuning · Changing Context Is Not Changing the Model
same need → prompt / retrieval / fine-tuning → compare parameter
state
微调(Fine-tuning)是在已有模型参数上继续训练,使其适应特定任务、领域或行为。它真的改变参数,因此和提示、RAG
的推理时控制必须分开。Fine-tuning continues training from existing model parameters to adapt
a task, domain, or behavior. It truly changes parameters and must be
distinguished from inference-time prompting or RAG.
01 · 先从日常问题开始01 · Start with an everyday question
员工临时看一张操作卡,和参加两周培训后形成习惯——这两种改变会留在同一个地方吗?An employee reads a one-off job aid versus trains for two weeks. Do
those changes live in the same place?
Δ
02同一需求走三条轨,只看哪里动了参数Route One Need Three Ways and Watch Which Changes Parameters
选择路线并调样本数choose route + examples
参数变化Parameter change
—
toy Δθ
上线成本Upfront cost
—
relative units
回滚难度Rollback effort
—
relative level
03三条路线改变的是不同层The Three Routes Change Different Layers
把刚才的变化拆开Map the change to parts
01 · PROMPT
提示改输入Prompt changes input
每次请求带上任务、约束或示例,不改基础参数。Each request carries tasks, constraints, or examples without
changing base parameters.
02 · RAG
检索改上下文Retrieval changes context
在推理时加入外部证据,知识可更新,参数保持不变。External evidence enters at inference time; knowledge can update
while parameters stay fixed.
03 · FT
微调改参数Fine-tuning changes parameters
用适配数据继续训练,行为变化进入新版本权重。Continued training on adaptation data moves behavior into a new
parameter version.
04 · OPS
版本与回滚Version and rollback
改参数后必须保存基线、检查点和可切换版本。Parameter changes require a baseline, checkpoints, and switchable
versions.
04最小判别:有没有发生参数更新Minimal Test: Did Parameters Update?
fine-tune: θ′ = θ − ηg
prompt/RAG: θ′ = θ fine-tune: θ′ = θ − ηg
三条轨使用同一需求。提示和检索改变当前输入;只有微调根据适配样本计算更新并产生新参数版本。All routes receive the same need. Prompting and retrieval change
current input; only fine-tuning computes updates from adaptation
examples and creates a new parameter version.
同一业务需求Same business need固定比较起点fix the comparison
选择控制层Choose control layer输入、上下文或参数input, context, or parameters
比较成本回滚Compare cost and rollback按变化位置治理govern where change lives
05边界实验:零样本也叫微调Boundary Test: Call Zero Examples Fine-tuning
主动制造失败Create a failure
选择微调,却不给任何适配样本Choose fine-tuning with no adaptation examples
没有样本就没有更新方向;即使按钮写着“微调”,参数变化仍为
0。Without examples there is no update direction. A button labeled
fine-tune still produces zero parameter change.
尚未执行。先在上方改变主变量。Not run yet. Change the main variable above first.
✗ 把文档塞进提示就是微调。Putting documents in a prompt is fine-tuning.
→ ✓ 那是改变上下文,基础参数不变。That changes context; base parameters stay fixed.
✗ 微调后无需保留旧版本。The old version is unnecessary after fine-tuning.
→ ✓ 参数变化必须支持检查点与回滚。Parameter changes require checkpoints and
rollback.
✗ 任何需求都应优先微调。Fine-tune every new need first.
→ ✓ 先比较提示、检索、成本与更新频率。Compare prompting, retrieval, cost, and update frequency
first.
06 · 一句话带走06 · One line to keep
提示改输入,检索改上下文,微调才继续训练并改变参数;选路线时必须同时比较成本、更新频率与回滚责任。Prompting changes input, retrieval changes context, and fine-tuning
continues training to change parameters. Route selection must
include cost, update cadence, and rollback responsibility.