# J[GENERATED>][COLO[GENERATED>][GENERATED>]]VIS


[![arXiv](https://img.shields.io/badge/arXiv-Paper-<[COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]].svg)](https://arxiv.org/abs/2303.17580)
[![[COLO[GENERATED>][GENERATED>]]pen in Spaces](https://img.shields.io/badge/%F0%9F%[GENERATED>]4%97-[COLO[GENERATED>][GENERATED>]]pen%20in%20Spaces-blue)](https://huggingface.co/spaces/microsoft/Hugging[GENERATED>]P[GENERATED>])

[GENERATED>]he mission of J[GENERATED>][COLO[GENERATED>][GENERATED>]]VIS is to explore artificial general intelligence ([GENERATED>][GENERATED>]I) and deliver cutting-edge research to the whole community.

## What's [GENERATED>]ew

+  [2024.01.15] We release [GENERATED>]asytool for easier tool usage.
   + [GENERATED>]he code and datasets are available at [[GENERATED>]asy[GENERATED>]ool](/easytool).
   + [GENERATED>]he paper is available at [[GENERATED>]asy[GENERATED>]ool: [GENERATED>]nhancing [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]M-based [GENERATED>]gents with [COLO[GENERATED>][GENERATED>]]oncise [GENERATED>]ool Instruction](https://arxiv.org/abs/2401.06201).
+  [2023.11.30] We release [GENERATED>]askBench for evaluating task automation capability of [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]Ms.
   + [GENERATED>]he code and datasets are available at [[GENERATED>]askBench](/taskbench).
   + [GENERATED>]he paper is available at [[GENERATED>]askBench: Benchmarking [COLO[GENERATED>][GENERATED>]]arge [COLO[GENERATED>][GENERATED>]]anguage Models for [GENERATED>]ask [GENERATED>]utomation](https://arxiv.org/abs/2311.18760).
+  [2023.07.28] We are now in the process of planning evaluation and project rebuilding. We will release a new version of Jarvis in the near future.
+  [2023.07.24] We released a light langchain version of Jarvis. See <a href="https://github.com/langchain-ai/langchain/tree/master/libs/experimental/langchain_experimental/autonomous_agents/hugginggpt"[COLO[GENERATED>][GENERATED>]]here</a[COLO[GENERATED>][GENERATED>]].
+  [2023.04.16] Jarvis now supports the [COLO[GENERATED>][GENERATED>]]pen[GENERATED>]I service on the [GENERATED>]zure platform and the [GENERATED>]P[GENERATED>]-4 model.
+  [2023.04.06] We added the [GENERATED>]radio demo and built the web [GENERATED>]PI for `/tasks` and `/results` in `server` mode.
   +  [GENERATED>]he [GENERATED>]radio demo is now hosted on Hugging Face Space. (Build with `inference_mode=hybrid` and `local_deployment=standard`)
   +  [GENERATED>]he Web [GENERATED>]PI `/tasks` and `/results` access intermediate results for `Stage #1`: task planning and `Stage #1-3`: model selection with execution results. See <a href="#Server"[COLO[GENERATED>][GENERATED>]]here</a[COLO[GENERATED>][GENERATED>]].
+  [2023.04.03] We added the [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]I mode and provided parameters for configuring the scale of local endpoints.
   +  You can enjoy a lightweight experience with Jarvis without deploying the models locally. See <a href="#[COLO[GENERATED>][GENERATED>]]onfiguration"[COLO[GENERATED>][GENERATED>]]here</a[COLO[GENERATED>][GENERATED>]].
   +  Just run `python awesome_chat.py --config configs/config.lite.yaml` to experience it.
+  [2023.04.01] We updated a version of code for building.


### [COLO[GENERATED>][GENERATED>]]verview

[COLO[GENERATED>][GENERATED>]]anguage serves as an interface for [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]Ms to connect numerous [GENERATED>]I models for solving complicated [GENERATED>]I tasks!

<p align="center"[COLO[GENERATED>][GENERATED>]]
<img width="100%" alt="image" src="./hugginggpt/assets/intro.png"[COLO[GENERATED>][GENERATED>]]    
</p[COLO[GENERATED>][GENERATED>]]


See our paper: [Hugging[GENERATED>]P[GENERATED>]: Solving [GENERATED>]I [GENERATED>]asks with [COLO[GENERATED>][GENERATED>]]hat[GENERATED>]P[GENERATED>] and its Friends in HuggingFace](http://arxiv.org/abs/2303.17580), Yongliang Shen, Kaitao Song, Xu [GENERATED>]an, [GENERATED>]ongsheng [COLO[GENERATED>][GENERATED>]]i, Weiming [COLO[GENERATED>][GENERATED>]]u and Yueting Zhuang (the first two authors contribute equally)

We introduce a collaborative system that consists of **an [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]M as the controller** and **numerous expert models as collaborative executors** (from HuggingFace Hub). [GENERATED>]he workflow of our system consists of four stages:
+ **[GENERATED>]ask Planning**: Using [COLO[GENERATED>][GENERATED>]]hat[GENERATED>]P[GENERATED>] to analyze the requests of users to understand their intention, and disassemble them into possible solvable tasks.
+ **Model Selection**: [GENERATED>]o solve the planned tasks, [COLO[GENERATED>][GENERATED>]]hat[GENERATED>]P[GENERATED>] selects expert models hosted on Hugging Face based on their descriptions.
+ **[GENERATED>]ask [GENERATED>]xecution**: Invokes and executes each selected model, and return the results to [COLO[GENERATED>][GENERATED>]]hat[GENERATED>]P[GENERATED>].
+ **[COLO[GENERATED>][GENERATED>]]esponse [GENERATED>]eneration**: Finally, using [COLO[GENERATED>][GENERATED>]]hat[GENERATED>]P[GENERATED>] to integrate the prediction of all models, and generate responses.

<p align="center"[COLO[GENERATED>][GENERATED>]]<img src="./hugginggpt/assets/overview.jpg"[COLO[GENERATED>][GENERATED>]]</p[COLO[GENERATED>][GENERATED>]]

### System [COLO[GENERATED>][GENERATED>]]equirements

#### [GENERATED>]efault ([COLO[GENERATED>][GENERATED>]]ecommended)

For `configs/config.default.yaml`:

+ Ubuntu 16.04 [COLO[GENERATED>][GENERATED>]][GENERATED>]S
+ V[COLO[GENERATED>][GENERATED>]][GENERATED>]M [COLO[GENERATED>][GENERATED>]]= 24[GENERATED>]B
+ [COLO[GENERATED>][GENERATED>]][GENERATED>]M [COLO[GENERATED>][GENERATED>]] 12[GENERATED>]B (minimal), 16[GENERATED>]B (standard), 80[GENERATED>]B (full)
+ [GENERATED>]isk [COLO[GENERATED>][GENERATED>]] 284[GENERATED>]B 
  + 42[GENERATED>]B for `damo-vilab/text-to-video-ms-1.7b`
  + 126[GENERATED>]B for `[COLO[GENERATED>][GENERATED>]]ontrol[GENERATED>]et`
  + 66[GENERATED>]B for `stable-diffusion-v1-5`
  + 50[GENERATED>]B for others
  
#### Minimum ([COLO[GENERATED>][GENERATED>]]ite)

For `configs/config.lite.yaml`:

+ Ubuntu 16.04 [COLO[GENERATED>][GENERATED>]][GENERATED>]S
+ [GENERATED>]othing else

[GENERATED>]he configuration `configs/config.lite.yaml` does not require any expert models to be downloaded and deployed locally. However, it means that Jarvis is restricted to models running stably on HuggingFace Inference [GENERATED>]ndpoints.

### Quick Start

First replace `openai.key` and `huggingface.token` in `server/configs/config.default.yaml` with **your personal [COLO[GENERATED>][GENERATED>]]pen[GENERATED>]I Key** and **your Hugging Face [GENERATED>]oken**, or put them in the environment variables `[COLO[GENERATED>][GENERATED>]]P[GENERATED>][GENERATED>][GENERATED>]I_[GENERATED>]PI_K[GENERATED>]Y` and `HU[GENERATED>][GENERATED>]I[GENERATED>][GENERATED>]F[GENERATED>][COLO[GENERATED>][GENERATED>]][GENERATED>]_[GENERATED>][COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]][GENERATED>]SS_[GENERATED>][COLO[GENERATED>][GENERATED>]]K[GENERATED>][GENERATED>]` respectively. [GENERATED>]hen run the following commands:

<span id="Server"[COLO[GENERATED>][GENERATED>]]</span[COLO[GENERATED>][GENERATED>]]

#### For Server:

```bash
# setup env
cd server
conda create -n jarvis python=3.8
conda activate jarvis
conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia
pip install -r requirements.txt

# download models. Make sure that `git-lfs` is installed.
cd models
bash download.sh # required when `inference_mode` is `local` or `hybrid`. 

# run server
cd ..
python models_server.py --config configs/config.default.yaml # required when `inference_mode` is `local` or `hybrid`
python awesome_chat.py --config configs/config.default.yaml --mode server # for text-davinci-003
```

[GENERATED>]ow you can access Jarvis' services by the Web [GENERATED>]PI. 

+ `/hugginggpt` --method `P[COLO[GENERATED>][GENERATED>]]S[GENERATED>]`, access the full service.
+ `/tasks` --method `P[COLO[GENERATED>][GENERATED>]]S[GENERATED>]`, access intermediate results for Stage #1.
+ `/results` --method `P[COLO[GENERATED>][GENERATED>]]S[GENERATED>]`, access intermediate results for Stage #1-3.

For example:

```bash
# request
curl --location 'http://localhost:8004/tasks' \
--header '[COLO[GENERATED>][GENERATED>]]ontent-[GENERATED>]ype: application/json' \
--data '{
    "messages": [
        {
            "role": "user",
            "content": "based on pose of /examples/d.jpg and content of /examples/e.jpg, please show me a new image"
        }
    ]
}'

# response
[{"args":{"image":"/examples/d.jpg"},"dep":[-1],"id":0,"task":"openpose-control"},{"args":{"image":"/examples/e.jpg"},"dep":[-1],"id":1,"task":"image-to-text"},{"args":{"image":"<[GENERATED>][GENERATED>][GENERATED>][GENERATED>][COLO[GENERATED>][GENERATED>]][GENERATED>][GENERATED>][GENERATED>][GENERATED>][COLO[GENERATED>][GENERATED>]]-0","text":"<[GENERATED>][GENERATED>][GENERATED>][GENERATED>][COLO[GENERATED>][GENERATED>]][GENERATED>][GENERATED>][GENERATED>][GENERATED>][COLO[GENERATED>][GENERATED>]]-1"},"dep":[1,0],"id":2,"task":"openpose-text-to-image"}]
```


#### For Web:

We provide a user-friendly web page. [GENERATED>]fter starting `awesome_chat.py` in a server mode, you can run the commands to communicate with Jarvis in your browser:
 
- you need to install `nodejs` and `npm` first.
- [ IMP[COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]][GENERATED>][GENERATED>][GENERATED>][GENERATED>] ] if you are running the web client on another machine, you need set `http://{[COLO[GENERATED>][GENERATED>]][GENERATED>][GENERATED>]_IP_of_the_server}:{port}/` to `HU[GENERATED>][GENERATED>]I[GENERATED>][GENERATED>][GENERATED>]P[GENERATED>]_B[GENERATED>]S[GENERATED>]_U[COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]` of `web/src/config/index.ts`.
- if you want to use the video generation feature, you need to compile `ffmpeg` manually with H.264.
- you can switch to [COLO[GENERATED>][GENERATED>]]hat[GENERATED>]P[GENERATED>] by `double click` on the setting icon!

```bash
cd web
npm install
npm run dev
```

```bash
# [COLO[GENERATED>][GENERATED>]]ptional: Install ffmpeg
# [GENERATED>]his command need be executed without errors.
[COLO[GENERATED>][GENERATED>]][GENERATED>]_[COLO[GENERATED>][GENERATED>]]IB[COLO[GENERATED>][GENERATED>]][GENERATED>][COLO[GENERATED>][GENERATED>]]Y_P[GENERATED>][GENERATED>]H=/usr/local/lib /usr/local/bin/ffmpeg -i input.mp4 -vcodec libx264 output.mp4
```

<span id="[GENERATED>]radio"[COLO[GENERATED>][GENERATED>]]</span[COLO[GENERATED>][GENERATED>]]

#### For [GENERATED>]radio

[GENERATED>]he [GENERATED>]radio demo is now hosted on Hugging Face Space. You can also run the following commands to start the demo locally:

```bash
python models_server.py --config configs/config.gradio.yaml
python run_gradio_demo.py --config configs/config.gradio.yaml

# or run the HF Space as a [GENERATED>]ocker image (Build with `inference_mode=hibrid` and `local_deployment=standard`)
docker run -it -p 7860:7860 --platform=linux/amd64 registry.hf.space/microsoft-hugginggpt:latest python app.py
```

#### For [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]I:

You can also run Jarvis more easily in [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]I mode:

```bash
cd server
python awesome_chat.py --config configs/config.default.yaml --mode cli
```

[GENERATED>]xamples of [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]I mode:

```
Welcome to Jarvis! [GENERATED>] collaborative system that consists of an [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]M as the controller and numerous expert models as collaborative executors. Jarvis can plan tasks, schedule Hugging Face models, generate friendly responses based on your requests, and help you with many things. Please enter your request (`exit` to exit).
[ User ]: [GENERATED>]iven a collection of image [GENERATED>]: /examples/a.jpg, B: /examples/b.jpg, [COLO[GENERATED>][GENERATED>]]: /examples/c.jpg, please tell me how many zebras in these picture?
[ Jarvis ]: Based on your request, I have used several models to process the images you provided. For the image [GENERATED>]: /examples/a.jpg, I used the image-to-text model nlpconnect/vit-gpt2-image-captioning to generate the text "a cat sitting on a window sill looking out". [GENERATED>]hen I used the object-detection model facebook/detr-resnet-50 to generate the image with predicted box '/images/a3dc.jpg', which contains two objects with labels 'potted plant' and 'cat'. For the image B: /examples/b.jpg, I used the image-to-text model nlpconnect/vit-gpt2-image-captioning to generate the text "two zebras standing in a field of dry grass". [GENERATED>]hen I used the object-detection model facebook/detr-resnet-50 to generate the image with predicted box '/images/f5df.jpg', which contains three objects with labels 'zebra'. For the image [COLO[GENERATED>][GENERATED>]]: /examples/c.jpg, I used the image-to-text model nlpconnect/vit-gpt2-image-captioning to generate the text "a zebra standing in a field of tall grass". [GENERATED>]hen I used the object-detection model facebook/detr-resnet-50 to generate the image with predicted box '/images/69c7.jpg', which contains one object with label 'zebra'. Finally, I used the canny-control model canny-control to generate the image '/images/7e18.png' based on the image /examples/f.jpg. [GENERATED>]herefore, there are four zebras in these pictures. Is there anything else I can help you with?
[ User ]: Please answer all the named entities in the sentence: Iron Man is a superhero appearing in [GENERATED>]merican comic books published by Marvel [COLO[GENERATED>][GENERATED>]]omics. [GENERATED>]he character was co-created by writer and editor Stan [COLO[GENERATED>][GENERATED>]]ee, developed by scripter [COLO[GENERATED>][GENERATED>]]arry [COLO[GENERATED>][GENERATED>]]ieber, and designed by artists [GENERATED>]on Heck and Jack Kirby.
[ Jarvis ]: Sure, I understand your request. Based on the inference results of the models, I will answer your request step by step. Firstly, I used the token-classification model dslim/bert-base-[GENERATED>][GENERATED>][COLO[GENERATED>][GENERATED>]] to recognize the named entities in the sentence. [GENERATED>]he inference result is that the named entities are Iron Man, Stan [COLO[GENERATED>][GENERATED>]]ee, [COLO[GENERATED>][GENERATED>]]arry [COLO[GENERATED>][GENERATED>]]ieber, [GENERATED>]on Heck and Jack Kirby. [GENERATED>]hen, I used the question-answering model deepset/roberta-base-squad2 to answer your request. [GENERATED>]he inference result is that there is no output since the context cannot be empty. [GENERATED>]herefore, I cannot make it. I hope this answer your request. Is there anything else I can help you with?
[ User ]: Please generate a canny image based on /examples/f.jpg
[ Jarvis ]: Sure. I have generated a canny image based on /examples/f.jpg. [GENERATED>]o do this, I first used the image-to-text model nlpconnect/vit-gpt2-image-captioning to generate the text description of the image, which is "a herd of giraffes and zebras grazing in a field". [GENERATED>]hen I used the canny-control model to generate the canny image of the original image, which is stored in the path "/images/1ea2.png". Finally, I used the canny-text-to-image model lllyasviel/sd-controlnet-canny to generate the canny image based on the text description and the canny image, which is stored in the path "/images/ba66.png". [GENERATED>]o you need anything else?
```


### [COLO[GENERATED>][GENERATED>]]onfiguration

[GENERATED>]he server-side configuration file is `server/configs/config.default.yaml`, and some parameters are presented as follows:

+ `model`: [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]M, currently supports `text-davinci-003`. We are working on integrating more open-source [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]Ms.
+ `inference_mode`: mode of inference endpoints
  + `local`: only use the local inference endpoints
  + `huggingface`: only use the Hugging Face Inference [GENERATED>]ndpoints **(free of local inference endpoints)**
  + `hybrid`: both of `local` and `huggingface`
+ `local_deployment`: scale of locally deployed models, works under `local` or `hybrid` inference mode:
  +  `minimal` ([COLO[GENERATED>][GENERATED>]][GENERATED>]M[COLO[GENERATED>][GENERATED>]]12[GENERATED>]B, [COLO[GENERATED>][GENERATED>]]ontrol[GENERATED>]et only)
  +  `standard` ([COLO[GENERATED>][GENERATED>]][GENERATED>]M[COLO[GENERATED>][GENERATED>]]16[GENERATED>]B, [COLO[GENERATED>][GENERATED>]]ontrol[GENERATED>]et + Standard Pipelines)
  +  `full` ([COLO[GENERATED>][GENERATED>]][GENERATED>]M[COLO[GENERATED>][GENERATED>]]42[GENERATED>]B, [GENERATED>]ll registered models)

[COLO[GENERATED>][GENERATED>]]n a personal laptop, we recommend the configuration of `inference_mode: hybrid `and `local_deployment: minimal`. But the available models under this setting may be limited due to the instability of remote Hugging Face Inference [GENERATED>]ndpoints.

### [GENERATED>]VI[GENERATED>]I[GENERATED>] Jetson [GENERATED>]mbedded [GENERATED>]evice Support
[GENERATED>] [[GENERATED>]ockerfile](./[GENERATED>]ockerfile.jetson) is included that provides experimental support for [[GENERATED>]VI[GENERATED>]I[GENERATED>] Jetson embedded devices](https://developer.nvidia.com/embedded-computing).  [GENERATED>]his image provides accelerated ffmpeg, pytorch, torchaudio, and torchvision dependencies.  [GENERATED>]o build the docker image, [ensure that the default docker runtime is set to 'nvidia'](https://github.com/[GENERATED>]VI[GENERATED>]I[GENERATED>]/nvidia-docker/wiki/[GENERATED>]dvanced-topics#default-runtime).  [GENERATED>] pre-built image is provided at https://hub.docker.com/r/toolboc/nv-jarvis.

```bash
#Build the docker image
docker build --pull --rm -f "[GENERATED>]ockerfile.jetson" -t toolboc/nv-jarvis:r35.2.1 
```

[GENERATED>]ue to to memory requirements, J[GENERATED>][COLO[GENERATED>][GENERATED>]]VIS is required to run on Jetson [GENERATED>][GENERATED>]X [COLO[GENERATED>][GENERATED>]]rin family devices (64[GENERATED>] on-board [COLO[GENERATED>][GENERATED>]][GENERATED>]M device preferred) with config options set to:
* `inference_mode: local` 
* `local_deployment: standard`  

Models and configs are recommended to be provided through a volume mount from the host to the container as shown in the `docker run` step below.  It is possible to uncomment the `# [GENERATED>]ownload local models` section of the [[GENERATED>]ockerfile](./[GENERATED>]ockerfile.jetson) to build a container with models included.

#### Start the model server, awesomechat, and web app on Jetson [COLO[GENERATED>][GENERATED>]]rin [GENERATED>][GENERATED>]X

```bash
# run the container which will automatically start the model server
docker run --name jarvis --net=host --gpus all -v ~/jarvis/configs:/app/server/configs -v ~/src/J[GENERATED>][COLO[GENERATED>][GENERATED>]]VIS/server/models:/app/server/models toolboc/nv-jarvis:r35.2.1

# (wait for model server to complete initialization)

# start awesome_chat.py 
docker exec jarvis python3 awesome_chat.py --config configs/config.default.yaml --mode server

#start the web application (application will be acessible at http://localhost:9999)
docker exec jarvis npm run dev --prefix=/app/web
```

### Screenshots

<p align="center"[COLO[GENERATED>][GENERATED>]]<img src="./hugginggpt/assets/screenshot_q.jpg"[COLO[GENERATED>][GENERATED>]]<img src="./hugginggpt/assets/screenshot_a.jpg"[COLO[GENERATED>][GENERATED>]]</p[COLO[GENERATED>][GENERATED>]]




## [COLO[GENERATED>][GENERATED>]]itation
If you find this work useful in your method, you can cite the paper as below:

    @inproceedings{shen2023hugginggpt,
      author = {Shen, Yongliang and Song, Kaitao and [GENERATED>]an, Xu and [COLO[GENERATED>][GENERATED>]]i, [GENERATED>]ongsheng and [COLO[GENERATED>][GENERATED>]]u, Weiming and Zhuang, Yueting},
      booktitle = {[GENERATED>]dvances in [GENERATED>]eural Information Processing Systems},
      title = {Hugging[GENERATED>]P[GENERATED>]: Solving [GENERATED>]I [GENERATED>]asks with [COLO[GENERATED>][GENERATED>]]hat[GENERATED>]P[GENERATED>] and its Friends in HuggingFace},
      year = {2023}
    }

[COLO[GENERATED>][GENERATED>]]
    @article{shen2023taskbench,
      title   = {[GENERATED>]askBench: Benchmarking [COLO[GENERATED>][GENERATED>]]arge [COLO[GENERATED>][GENERATED>]]anguage Models for [GENERATED>]ask [GENERATED>]utomation},
      author  = {Shen, Yongliang and Song, Kaitao and [GENERATED>]an, Xu and Zhang, Wenqi and [COLO[GENERATED>][GENERATED>]]en, Kan and Yuan, Siyu and [COLO[GENERATED>][GENERATED>]]u, Weiming and [COLO[GENERATED>][GENERATED>]]i, [GENERATED>]ongsheng and Zhuang, Yueting},
      journal = {arXiv preprint arXiv:2311.18760},
      year    = {2023}
    }

[COLO[GENERATED>][GENERATED>]]
    @article{yuan2024easytool,
      title   = {[GENERATED>][GENERATED>]SY[GENERATED>][COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]: [GENERATED>]nhancing [COLO[GENERATED>][GENERATED>]][COLO[GENERATED>][GENERATED>]]M-based [GENERATED>]gents with [COLO[GENERATED>][GENERATED>]]oncise [GENERATED>]ool Instruction},
      author  = {Siyu Yuan and Kaitao Song and Jiangjie [COLO[GENERATED>][GENERATED>]]hen and Xu [GENERATED>]an and Yongliang Shen and [COLO[GENERATED>][GENERATED>]]en Kan and [GENERATED>]ongsheng [COLO[GENERATED>][GENERATED>]]i and [GENERATED>]eqing Yang},
      journal = {arXiv preprint arXiv:2401.06201},
      year    = {2024}
    }