diff --git "a/docs/source/Instruction/\346\216\250\347\220\206\345\222\214\351\203\250\347\275\262.md" "b/docs/source/Instruction/\346\216\250\347\220\206\345\222\214\351\203\250\347\275\262.md"
index 76a574b95..ca887b619 100644
--- "a/docs/source/Instruction/\346\216\250\347\220\206\345\222\214\351\203\250\347\275\262.md"
+++ "b/docs/source/Instruction/\346\216\250\347\220\206\345\222\214\351\203\250\347\275\262.md"
@@ -15,7 +15,7 @@ SWIFT支持以命令行、Python代码和界面方式进行推理和部署：
 - `single-line`命令 切换到单行模式
 - `clear`命令 清除history
 - `exit`命令 退出
-- 如果query中带有多模态数据，添加<image>/<video>/<audio>等标签，例如输入`<image>What is in the image?`，即可在接下来输入图片地址
+- 如果query中带有多模态数据，添加`<image>/<video>/<audio>`等标签，例如输入`<image>What is in the image?`，即可在接下来输入图片地址
 
 ## 推理加速后端
 
diff --git a/docs/source_en/Instruction/Inference-and-deployment.md b/docs/source_en/Instruction/Inference-and-deployment.md
index 1229ba359..c7c0fa2e0 100644
--- a/docs/source_en/Instruction/Inference-and-deployment.md
+++ b/docs/source_en/Instruction/Inference-and-deployment.md
@@ -16,7 +16,7 @@ The command line inference can be referred to via the link provided in the secon
 - The `single-line` command switches to single-line mode.
 - The `clear` command clears the history.
 - The `exit` command exits the application.
-If the query involves multimodal data, add tags like <image>/<video>/<audio>. For example, input `<image>What is in the image?`, and you can then input the image address.
+If the query involves multimodal data, add tags like `<image>/<video>/<audio>`. For example, input `<image>What is in the image?`, and you can then input the image address.
 
 ## Inference Acceleration Backend
 You can perform inference and deployment using `swift infer/deploy`. Currently, SWIFT supports three inference frameworks: pt (native torch), vLLM, and LMDeploy. You can switch between them using `--infer_backend pt/vllm/lmdeploy`. Apart from pt, both vLLM and LMDeploy have their own model support ranges. Please refer to their official documentation to verify availability and prevent runtime errors.