Skip to content

I didn't put the tool in the training call, why does the model after Qwen2-14b-instruct training have the fields <translation> and </translation> when reasoning? Is it because the Qwen pre-training process adds a lot of similar data training? #49

Open

Description

@shaoxiong007

Question: 截至今日,我国的联合国旅游组织“最佳旅游乡村”总数达到多少个,2024年我国有几个乡村入选“最佳旅游乡村”名单,分别是哪几个?
<|im_end|>
<|im_start|>assistant

我需要先查询全国的“最佳旅游乡村”总数,然后查询2024年入选的乡村名单及数量。但是,由于2024年还未到,所以可能会查不到具体的入选名单和村庄数量。我将首先尝试搜索现有“最佳旅游乡村”的总数。

Почемуда, я не могу выполнить поиск за 2024 год, так как это находится в будущем и информации об этом пока нет. Я начну с поиска текущего количества отмеченных "лучших乡村旅游" в Китае.

< translation>
然而,我不能针对2024年进行搜索,因为那是在未来没有相关信息。我将首先查询当前我国获得联合国“最佳旅游乡村”的总数量。
</ translation>

<tool_call>
{"name": "web_search", "arguments": {"query": "我国现有联合国旅游组织最佳旅游乡村总数"}}
</tool_call><|im_end|>

<|im_start|>user

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions