紫东太初

Zidong Taichu is a fully modal large model jointly launched by the Institute of Automation of the Chinese Academy of Sciences and the Wuhan Institute of Artificial Intelligence. It is an upgraded version 2.0 based on the multi-modal large model “Zidong Taichu 1.0” with billions of parameters. The Zidong Taichu model supports comprehensive Q&A tasks such as multiple rounds of question and answer, text creation, image generation, 3D understanding, signal analysis, etc. It has strong cognitive, comprehension, and creative abilities and can bring a brand new interactive experience.
The main functions of Zidong Taichu
1、 Image capability
Image description: Based on user uploaded image materials, accurately understand and answer image recognition questions
Object detection: Supports object detection tasks for diverse categories, capable of determining target types, quantities, and corresponding location information
Image retrieval: Based on a massive library of high-quality image materials, users can search for highly relevant and exquisite image materials
Image generation: Based on user instructions and demands, generate corresponding exquisite images, and can modify descriptions to fine tune image content
Text recognition: Based on image content, supports multi scene, multilingual, and high-precision text detection and recognition services
2、 Language ability
Chinese Q&A: Accurately understand the context of user input questions and be able to provide accurate knowledge-based Q&A
Text Continuation: Automatic Continuation of Rich Story Content Guided by User Input
Text creation: Accurately understand the user’s input intention and generate coherent and logically coherent text content
Title generation: Based on understanding the article or long text, quickly aggregate and generate concise and summarized title copy
Grammar analysis: Accurately understand and analyze the grammar of Chinese and English sentences, remind grammar errors and make corrections, etc
Machine translation: Helps users translate various types of text materials, including Chinese English translation, classical Chinese and vernacular Chinese translation, etc
Ancient poetry creation: improvise poetry, quatrains, etc. based on the theme or guiding content provided by the user
Code understanding: Understand the vast majority of programming languages, algorithms, and data structures such as C, Python, and JAVA, and quickly provide the necessary answers
Code writing: can help users quickly write simple code snippets, such as functions, classes, or loops
Mathematical calculation: It can handle both conventional mathematical calculation problems and mathematical application problems such as the chicken rabbit cage mathematical application problems recorded in the Sun Tzu Mathematical Classic
Logical reasoning: supports handling complex logical reasoning problems, including scientific reasoning, common sense reasoning, spatiotemporal reasoning, etc
3、 Video capability
Video Description: Based on user uploaded video materials, accurately understand and answer questions related to video recognition and description
Video retrieval: Based on a massive library of high-quality video materials, users can search for highly relevant and exquisite video materials
Video Q&A: Based on user uploaded video materials, accurately understand and answer video related questions, while supporting contextual information understanding and multiple rounds of Q&A
4、 Music ability
Music generation: Controllably generate high fidelity music through given text prompts, and support improvisation of music with various styles and instruments for performance
Music multimodal Q&A: Based on the understanding of music materials uploaded by users, relevant multimodal Q&A tasks can be completed
5、 Audio capability
Audio authentication: Zidong Taichu can determine whether the current audio is spoken by a real person or synthesized by a machine
Audio Event Classification: It can detect the types of sound events contained in the current audio and currently supports 11 types of single sound events and mixed sound events
Speech recognition: It can quickly and accurately recognize speech as text, supporting multiple scenarios such as mobile application voice interaction and speech content analysis
Speech synthesis: Provides highly anthropomorphic, smooth and natural speech synthesis services to meet the needs of various scenarios such as text reading and speech broadcasting
6、 3D capability
3D scene description: Zidong Taichu 2.0 has the ability to understand 3D scenes and perceive objects based on point cloud data
7、 Signal capability
Signal recognition: supports radar signal identification and knowledge exchange, and can quickly grasp the basic sources and parameters of signals with the help of models
How to use Zidong Taichu
Visit the official website of Zidong Taichu (taichu web. ia. ac. cn) and click on the conversation experience
Log in/register your account, and once successfully applied for use, it will automatically redirect to the dialogue interface
Enter your question or enter a slash to select the recommended prompt command (you can also choose the built-in example to view), and then click send
Zidong Taichu will intelligently answer your questions
Frequently asked questions
How can Zidong Taichu apply for a trial?
Visit the official website of Zidong Taichu, click on registration on the login interface, enter “username”, “nickname”, “password”, “phone number” and other information to submit a registration account application. After waiting for the backend approval, you can use it for free.
Does Zidong Taichu support uploading files?
Zidong Taichu supports users to upload files of images, videos, point clouds, audio, music, and signals, and can conduct targeted dialogue and Q&A.
Has Zidong Taichu been registered through generative AI?
The Zidong Taichu model has been registered in the first batch of August 2023 through the Interim Measures for the Management of Generative Artificial Intelligence Services, and can be officially launched to provide services to the public.