> ## Documentation Index
> Fetch the complete documentation index at: https://langchain-zh.cn/llms.txt
> Use this file to discover all available pages before exploring further.

# 如何在界面中筛选实验

LangSmith 允许您根据反馈分数和元数据筛选过往实验，从而轻松找到您关心的实验。

## 背景：为实验添加元数据

在 SDK 中运行实验时，您可以附加元数据，以便在界面中更轻松地进行筛选。如果您在运行实验时已明确想要深入分析的维度，这会非常有用。

在我们的示例中，我们将围绕使用的模型、模型提供商以及已知的提示 ID 为实验附加元数据：

```python theme={"theme":{"light":"catppuccin-latte","dark":"catppuccin-mocha"}}
models = {
    "openai-gpt-4.1": ChatOpenAI(model="gpt-4.1", temperature=0),
    "openai-gpt-4.1-mini": ChatOpenAI(model="gpt-4.1-mini", temperature=0),
    "anthropic-claude-3-sonnet-20240229": ChatAnthropic(temperature=0, model_name="claude-3-sonnet-20240229")
}

prompts = {
    "singleminded": "始终用单词 banana 回答问题。",
    "fruitminded": "始终在回答中讨论水果。",
    "basic": "你是一个聊天机器人。"
}

def answer_evaluator(run, example) -> dict:
    llm = ChatOpenAI(model="gpt-4.1", temperature=0)
    answer_grader = hub.pull("langchain-ai/rag-answer-vs-reference") | llm
    score = answer_grader.invoke(
        {
            "question": example.inputs["question"],
            "correct_answer": example.outputs["answer"],
            "student_answer": run.outputs,
        }
    )
    return {"key": "correctness", "score": score["Score"]}

dataset_name = "可筛选数据集"

for model_type, model in models.items():
    for prompt_type, prompt in prompts.items():
        def predict(example):
            return model.invoke(
                [("system", prompt), ("user", example["question"])]
            )

        model_provider = model_type.split("-")[0]
        model_name = model_type[len(model_provider) + 1:]

        evaluate(
            predict,
            data=dataset_name,
            evaluators=[answer_evaluator],
            # 在此处添加元数据！！
            metadata={
                "model_provider": model_provider,
                "model_name": model_name,
                "prompt_id": prompt_type
            }
        )
```

## 在界面中筛选实验

在界面中，默认会显示所有已运行的实验。

<img src="https://mintcdn.com/hhh-8c10bf0c/lQoj_T05pUgIcyPg/langsmith/images/filter-all-experiments.png?fit=max&auto=format&n=lQoj_T05pUgIcyPg&q=85&s=0dd405e3b9ff0704ca2fb2801b603dc0" alt="筛选所有实验" width="2900" height="1370" data-path="langsmith/images/filter-all-experiments.png" />

例如，如果我们偏好 OpenAI 模型，可以轻松筛选并首先查看仅限 OpenAI 模型的分数：

<img src="https://mintcdn.com/hhh-8c10bf0c/lQoj_T05pUgIcyPg/langsmith/images/filter-openai.png?fit=max&auto=format&n=lQoj_T05pUgIcyPg&q=85&s=332591a655622c5e2e3a70c15151984a" alt="筛选 OpenAI" width="2910" height="1130" data-path="langsmith/images/filter-openai.png" />

我们可以叠加筛选条件，从而过滤掉正确性分数较低的实验，确保只比较相关的实验：

<img src="https://mintcdn.com/hhh-8c10bf0c/lQoj_T05pUgIcyPg/langsmith/images/filter-feedback.png?fit=max&auto=format&n=lQoj_T05pUgIcyPg&q=85&s=cbac36cf43cdd18023b94af297158d71" alt="筛选反馈" width="2912" height="826" data-path="langsmith/images/filter-feedback.png" />

最后，我们可以清除并重置筛选条件。例如，如果我们发现 `singleminded` 提示明显胜出，可以更改筛选设置，查看其他模型提供商的模型是否也能与之良好配合：

<img src="https://mintcdn.com/hhh-8c10bf0c/lQoj_T05pUgIcyPg/langsmith/images/filter-singleminded.png?fit=max&auto=format&n=lQoj_T05pUgIcyPg&q=85&s=939e2aa026de693cf77639c8cfcdb86b" alt="筛选 singleminded" width="2904" height="832" data-path="langsmith/images/filter-singleminded.png" />

***

<div className="source-links">
  <Callout icon="edit">
    [Edit this page on GitHub](https://github.com/langchain-ai/docs/edit/main/src/i18n\zh-CN\langsmith\filter-experiments-ui.mdx) or [file an issue](https://github.com/langchain-ai/docs/issues/new/choose).
  </Callout>

  <Callout icon="terminal-2">
    [Connect these docs](/use-these-docs) to Claude, VSCode, and more via MCP for real-time answers.
  </Callout>
</div>