Home Technology peripherals AI China's first self-developed MoE multi-modal large model reveals Tencent's mixed-element multi-modal understanding

China's first self-developed MoE multi-modal large model reveals Tencent's mixed-element multi-modal understanding

Aug 22, 2024 pm 10:38 PM
industry Tencent Yuanbao

Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

The AIxiv column is a column where this site publishes academic and technical content. In the past few years, the AIxiv column of this site has received more than 2,000 reports, covering top laboratories from major universities and companies around the world, effectively promoting academic exchanges and dissemination. If you have excellent work that you want to share, please feel free to contribute or contact us for reporting. Submission email: liyazhou@jiqizhixin.com; zhaoyunfeng@jiqizhixin.com


Large-scale language models represented by GPT herald the dawn of general artificial intelligence in the digital cognitive space. These models demonstrate powerful understanding and reasoning capabilities by processing and generating natural language, and have shown broad application prospects in multiple fields. Whether in content generation, automated customer service, productivity tools, AI search, or in fields such as education and medical care, large-scale language models are constantly promoting the advancement of technology and the popularization of applications.

However, to promote general artificial intelligence to explore the physical world, the first step is to solve the problem of visual understanding, that is, multi-modal understanding of large models. Multimodal understanding enables AI to more fully understand and interact with the world by acquiring and processing information through multiple senses, just like humans. Breakthroughs in this field will enable artificial intelligence to make greater progress in robotics, autonomous driving, etc., and truly realize the leap from the digital world to the physical world.

GPT-4V was released in June last year, but compared to large language models, the development of multi-modal understanding models appears to be slower, especially in the Chinese field. In addition, unlike the technical route and selection of large language models that are relatively certain, the industry has not yet fully reached a consensus on the architecture and selection of training methods for multi-modal models.

Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

                                                                                                                                                                                             dollars dollars dollars dollars dollars dollars State-of-the-art understanding of large models. The model has been innovative and deeply optimized in terms of architecture, training methods and data processing, significantly improving its performance and supporting the understanding of images with any aspect ratio and up to 7K resolution. Unlike most multi-modal models that are mainly tuned in open source benchmarks, Tencent's hybrid multi-modal model pays more attention to the versatility, practicality and reliability of the model, and has rich multi-modal scene understanding capabilities. In the recently released Chinese multi-modal large model SuperCLUE-V benchmark evaluation (August 2024), Tencent Hunyuan ranked first in the country, surpassing multiple mainstream closed-source models.

Method introduction: MoE architecture

Tencent’s large mixed language model is the first in China to adopt the mixed expert model (MoE) architecture. The overall performance of the model is 50% higher than the previous generation, and some Chinese capabilities It has tied up with GPT-4o, and has greatly improved its performance in answering "current" questions, as well as in mathematics, reasoning and other abilities. As early as the beginning of this year, Tencent Hunyuan applied this model to Tencent Yuanbao.
Tencent Hunyuan believes that the MoE architecture that can solve a large number of general tasks is also the best choice for multi-modal understanding scenarios. MoE can be better compatible with more modalities and tasks, ensuring that different modalities and tasks are mutually reinforcing rather than competitive.

Relying on the capabilities of Tencent Hunyuan's large language model, Tencent Hunyuan has launched a large multi-modal understanding model based on MoE architecture. It has made innovations and in-depth optimizations in terms of architecture, training methods and data processing, and its performance has been significantly improved. promote. This is also the first multi-modal large model based on MoE architecture in China.

模 Tencent mixed element multi -modal model architecture schematic diagram

Simple and large -scale
In addition to using the MOE architecture, the design of the Tencent mixed element multi -mode model also follows simple and reasonable , Principles of scalability:

Support native arbitrary resolutions: Compared with the industry’s mainstream fixed resolution or cut subgraph methods, Tencent’s hybrid multi-modal model can process native images of any resolution. Implemented the first multi-modal model to support image understanding with resolutions exceeding 7K and any aspect ratio (e.g. 16:1, see example below).
  • Using a simple MLP adapter: Compared with the previous mainstream Q-former adapter, the MLP adapter has less loss during information transmission.
This simple design makes it easier to expand and scale models and data.

SuperClue-V ranks first in the domestic list
In August 2024, SuperCLUE released the multi-modal understanding evaluation list for the first time - SuperClue-V.
The SuperCLUE-V benchmark includes two general directions: basic capabilities and application capabilities. It evaluates multi-modal large models in the form of open questions, including 8 first-level dimensions and 30 second-level dimensions.

In this evaluation, the Hunyuan multi-modal understanding system hunyuan-vision achieved a score of 71.95, second only to GPT-4o. In terms of multi-modal applications, hunyuan-vision is ahead of Claude3.5-Sonnet and Gemini-1.5-Pro. Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

It is worth noting that previous multi-modal assessments in the industry mostly focused on English proficiency, and most of the assessment questions were multiple-choice or true-false questions. The SuperCLUE-V evaluation focuses more on Chinese proficiency evaluation and focuses on users’ real problems. In addition, since this is the first release, overfitting has not yet occurred.

Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

Tencent Hunyuan Graphics and Text Large Model shows good performance in multiple dimensions such as general scenes, image OCR recognition and understanding, and Chinese element understanding and reasoning, and also reflects the potential of the model in future applications. Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understandingAimed at general application scenarios

The mixed-element multi-modal understanding model is optimized for general scenarios and massive applications, and has accumulated tens of millions of related question and answer corpus, covering basic image understanding, content creation, It can be used in many scenarios such as reasoning analysis, knowledge question and answer, OCR document analysis, and subject answering. The following are some typical application examples.

Here are more typical examples: Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

Convert an image into a text table:

Explain a piece of code: Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

Analyze a bill: Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

Description Picture content: Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

Do math problems: Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

Analyze based on the content of the picture: Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

Help you write copy: Chinas first self-developed MoE multi-modal large model reveals Tencents mixed-element multi-modal understanding

現在、Tencent の Hunyuan マルチモーダル理解大規模モデルは、AI アシスタント製品である Tencent Yuanbao でリリースされており、Tencent Cloud を通じて企業と個人の開発者に公開されています。

テンセント元宝アドレス: https://yuanbao.tencent.com/chat

The above is the detailed content of China's first self-developed MoE multi-modal large model reveals Tencent's mixed-element multi-modal understanding. For more information, please follow other related articles on the PHP Chinese website!

Statement of this Website
The content of this article is voluntarily contributed by netizens, and the copyright belongs to the original author. This site does not assume corresponding legal responsibility. If you find any content suspected of plagiarism or infringement, please contact admin@php.cn

Hot AI Tools

Undresser.AI Undress

Undresser.AI Undress

AI-powered app for creating realistic nude photos

AI Clothes Remover

AI Clothes Remover

Online AI tool for removing clothes from photos.

Undress AI Tool

Undress AI Tool

Undress images for free

Clothoff.io

Clothoff.io

AI clothes remover

AI Hentai Generator

AI Hentai Generator

Generate AI Hentai for free.

Hot Article

R.E.P.O. Energy Crystals Explained and What They Do (Yellow Crystal)
1 months ago By 尊渡假赌尊渡假赌尊渡假赌
R.E.P.O. Best Graphic Settings
1 months ago By 尊渡假赌尊渡假赌尊渡假赌
Will R.E.P.O. Have Crossplay?
1 months ago By 尊渡假赌尊渡假赌尊渡假赌

Hot Tools

Notepad++7.3.1

Notepad++7.3.1

Easy-to-use and free code editor

SublimeText3 Chinese version

SublimeText3 Chinese version

Chinese version, very easy to use

Zend Studio 13.0.1

Zend Studio 13.0.1

Powerful PHP integrated development environment

Dreamweaver CS6

Dreamweaver CS6

Visual web development tools

SublimeText3 Mac version

SublimeText3 Mac version

God-level code editing software (SublimeText3)

DeepMind robot plays table tennis, and its forehand and backhand slip into the air, completely defeating human beginners DeepMind robot plays table tennis, and its forehand and backhand slip into the air, completely defeating human beginners Aug 09, 2024 pm 04:01 PM

But maybe he can’t defeat the old man in the park? The Paris Olympic Games are in full swing, and table tennis has attracted much attention. At the same time, robots have also made new breakthroughs in playing table tennis. Just now, DeepMind proposed the first learning robot agent that can reach the level of human amateur players in competitive table tennis. Paper address: https://arxiv.org/pdf/2408.03906 How good is the DeepMind robot at playing table tennis? Probably on par with human amateur players: both forehand and backhand: the opponent uses a variety of playing styles, and the robot can also withstand: receiving serves with different spins: However, the intensity of the game does not seem to be as intense as the old man in the park. For robots, table tennis

The first mechanical claw! Yuanluobao appeared at the 2024 World Robot Conference and released the first chess robot that can enter the home The first mechanical claw! Yuanluobao appeared at the 2024 World Robot Conference and released the first chess robot that can enter the home Aug 21, 2024 pm 07:33 PM

On August 21, the 2024 World Robot Conference was grandly held in Beijing. SenseTime's home robot brand "Yuanluobot SenseRobot" has unveiled its entire family of products, and recently released the Yuanluobot AI chess-playing robot - Chess Professional Edition (hereinafter referred to as "Yuanluobot SenseRobot"), becoming the world's first A chess robot for the home. As the third chess-playing robot product of Yuanluobo, the new Guoxiang robot has undergone a large number of special technical upgrades and innovations in AI and engineering machinery. For the first time, it has realized the ability to pick up three-dimensional chess pieces through mechanical claws on a home robot, and perform human-machine Functions such as chess playing, everyone playing chess, notation review, etc.

Claude has become lazy too! Netizen: Learn to give yourself a holiday Claude has become lazy too! Netizen: Learn to give yourself a holiday Sep 02, 2024 pm 01:56 PM

The start of school is about to begin, and it’s not just the students who are about to start the new semester who should take care of themselves, but also the large AI models. Some time ago, Reddit was filled with netizens complaining that Claude was getting lazy. "Its level has dropped a lot, it often pauses, and even the output becomes very short. In the first week of release, it could translate a full 4-page document at once, but now it can't even output half a page!" https:// www.reddit.com/r/ClaudeAI/comments/1by8rw8/something_just_feels_wrong_with_claude_in_the/ in a post titled "Totally disappointed with Claude", full of

At the World Robot Conference, this domestic robot carrying 'the hope of future elderly care' was surrounded At the World Robot Conference, this domestic robot carrying 'the hope of future elderly care' was surrounded Aug 22, 2024 pm 10:35 PM

At the World Robot Conference being held in Beijing, the display of humanoid robots has become the absolute focus of the scene. At the Stardust Intelligent booth, the AI ​​robot assistant S1 performed three major performances of dulcimer, martial arts, and calligraphy in one exhibition area, capable of both literary and martial arts. , attracted a large number of professional audiences and media. The elegant playing on the elastic strings allows the S1 to demonstrate fine operation and absolute control with speed, strength and precision. CCTV News conducted a special report on the imitation learning and intelligent control behind "Calligraphy". Company founder Lai Jie explained that behind the silky movements, the hardware side pursues the best force control and the most human-like body indicators (speed, load) etc.), but on the AI ​​side, the real movement data of people is collected, allowing the robot to become stronger when it encounters a strong situation and learn to evolve quickly. And agile

ACL 2024 Awards Announced: One of the Best Papers on Oracle Deciphering by HuaTech, GloVe Time Test Award ACL 2024 Awards Announced: One of the Best Papers on Oracle Deciphering by HuaTech, GloVe Time Test Award Aug 15, 2024 pm 04:37 PM

At this ACL conference, contributors have gained a lot. The six-day ACL2024 is being held in Bangkok, Thailand. ACL is the top international conference in the field of computational linguistics and natural language processing. It is organized by the International Association for Computational Linguistics and is held annually. ACL has always ranked first in academic influence in the field of NLP, and it is also a CCF-A recommended conference. This year's ACL conference is the 62nd and has received more than 400 cutting-edge works in the field of NLP. Yesterday afternoon, the conference announced the best paper and other awards. This time, there are 7 Best Paper Awards (two unpublished), 1 Best Theme Paper Award, and 35 Outstanding Paper Awards. The conference also awarded 3 Resource Paper Awards (ResourceAward) and Social Impact Award (

Hongmeng Smart Travel S9 and full-scenario new product launch conference, a number of blockbuster new products were released together Hongmeng Smart Travel S9 and full-scenario new product launch conference, a number of blockbuster new products were released together Aug 08, 2024 am 07:02 AM

This afternoon, Hongmeng Zhixing officially welcomed new brands and new cars. On August 6, Huawei held the Hongmeng Smart Xingxing S9 and Huawei full-scenario new product launch conference, bringing the panoramic smart flagship sedan Xiangjie S9, the new M7Pro and Huawei novaFlip, MatePad Pro 12.2 inches, the new MatePad Air, Huawei Bisheng With many new all-scenario smart products including the laser printer X1 series, FreeBuds6i, WATCHFIT3 and smart screen S5Pro, from smart travel, smart office to smart wear, Huawei continues to build a full-scenario smart ecosystem to bring consumers a smart experience of the Internet of Everything. Hongmeng Zhixing: In-depth empowerment to promote the upgrading of the smart car industry Huawei joins hands with Chinese automotive industry partners to provide

Li Feifei's team proposed ReKep to give robots spatial intelligence and integrate GPT-4o Li Feifei's team proposed ReKep to give robots spatial intelligence and integrate GPT-4o Sep 03, 2024 pm 05:18 PM

Deep integration of vision and robot learning. When two robot hands work together smoothly to fold clothes, pour tea, and pack shoes, coupled with the 1X humanoid robot NEO that has been making headlines recently, you may have a feeling: we seem to be entering the age of robots. In fact, these silky movements are the product of advanced robotic technology + exquisite frame design + multi-modal large models. We know that useful robots often require complex and exquisite interactions with the environment, and the environment can be represented as constraints in the spatial and temporal domains. For example, if you want a robot to pour tea, the robot first needs to grasp the handle of the teapot and keep it upright without spilling the tea, then move it smoothly until the mouth of the pot is aligned with the mouth of the cup, and then tilt the teapot at a certain angle. . this

Distributed Artificial Intelligence Conference DAI 2024 Call for Papers: Agent Day, Richard Sutton, the father of reinforcement learning, will attend! Yan Shuicheng, Sergey Levine and DeepMind scientists will give keynote speeches Distributed Artificial Intelligence Conference DAI 2024 Call for Papers: Agent Day, Richard Sutton, the father of reinforcement learning, will attend! Yan Shuicheng, Sergey Levine and DeepMind scientists will give keynote speeches Aug 22, 2024 pm 08:02 PM

Conference Introduction With the rapid development of science and technology, artificial intelligence has become an important force in promoting social progress. In this era, we are fortunate to witness and participate in the innovation and application of Distributed Artificial Intelligence (DAI). Distributed artificial intelligence is an important branch of the field of artificial intelligence, which has attracted more and more attention in recent years. Agents based on large language models (LLM) have suddenly emerged. By combining the powerful language understanding and generation capabilities of large models, they have shown great potential in natural language interaction, knowledge reasoning, task planning, etc. AIAgent is taking over the big language model and has become a hot topic in the current AI circle. Au

See all articles