Daimon Robotics Launches Global First Tactile-Grounded World Model Daimon-TWM

[](https://m.36kr.com/usercenter/basicinfo)[Account Settings](https://m.36kr.com/usercenter/account-password "Account Settings")[My Follows](https://m.36kr.com/usercenter/follow "My Follows")[My Favorites](https://m.36kr.com/usercenter/favorite "My Favorites")[Applied Projects](https://m.36kr.com/seek-report-new?tabKey=2 "Applied Projects")[Log Out](https://m.36kr.com/p/3934449071406212 "Log Out")

36Kr Auto Digital Time氪 Future Consumption Intelligent Emergence Future Cities Start Power on 36Kr Outbound 36Kr Research Institute TIDE 36Kr Enterprise Review 36Kr Finance Career Bonus 36Carbon Post浪 Institute Dark Waves Hard Kr Kr睿 Institute Media Brands
Company Accounts Enterprise Review 36Kr Research Institute 36Kr Innovation Consulting Company Services
Core Services City Window Government Services
Venture Release LP Source Plan VClub VClub Investment Institution Database Investment Institution Job Recommendations Investor Certification Investor Services
Project Recommendations 36Kr Pro Venture Krang Enterprise Entry Entrepreneur Services
Venture Platform
AI Evaluation Network

[](https://36kr.com/topics/799051634713857)[](http://www.bjjubao.org.cn/node_1606.html)[](ttps://36kr.com/topics/799051634713857)

[](https://m.36kr.com/)

Project Recommendations

I Want to Join

City Cooperation

Ant Group, leading a rare embodied team

Investment Circle_·_ 2026年08月11日 11:00

Building the world's first tactile-driven physical interaction brain.

Ant Group

Equity Financing Zhejiang Province _2000-10_

Global leading financial technology open platform

36Kr Report Finance High-Tech Enterprises

I Want to Contact

[](https://pitchhub.36kr.com/project/2144768335758592)

Today (August 11), Dai Meng Robotics announced the completion of hundreds of millions of yuan strategic round funding, with Ant Group leading the investment, and existing shareholders over-subscribing and adding more capital.

The fundraising process was rapid. Just two months ago, Dai Meng had just completed a 100 million yuan A round funding.

The investor lineup is even more impressive. Along with this round's Ant Group, Dai Meng has already gathered China Merchants Venture Capital, Lenovo Venture Capital, Huichuan Industrial Investment, China Mobile, China Telecom and a number of other top-tier industrial capital investors.

Why are the industrial giants flocking to invest in Dai Meng?

The answer lies in the shift of the focus of the embodied intelligence competition.

The next stage of the competition has moved from 'whether there is touch' to 'whether it can truly understand and use touch', and whether it can convert tactile input into tactile intelligence has become a decisive factor.

Industry research has already validated that introducing tactile data can significantly improve the success rate of contact-related tasks and data utilization efficiency, driving physical AI to outperform pure visual solutions in scaling curves.

Not long ago, the 'AI mother' Li Fei-Fei team proposed in the T-Rex paper that touch is not a subordinate of vision, but an independent, high-frequency physical perception channel. Simply adding tactile signals as additional inputs to existing models not only fails to fully enhance performance, but can even lead to a significant drop in task success rate.

If the T-Rex systemically solved the 'how to correctly use touch' problem, Dai Meng has taken it a step further by advancing touch from 'post-factum feedback' to 'pre-factum simulation'.

Dai Meng not only pioneered the VTLA model architecture (Vision-Tactile-Language-Action) in the industry, integrating touch natively into perception, decision-making, and action generation; but also developed the world's first 'physical interaction brain' that drives robotic dexterous operation, for the first time integrating physical cognition, simulation decision-making, and instant control into a unified model framework.

A tactile intelligence revolution driven by the 'physical interaction brain' is currently taking place, and Dai Meng is precisely at the center of this revolutionary storm.

From CMU to Hong Kong University of Science and Technology

Global rare tactile intelligence full-stack team

A truly driving 'physical interaction brain' for robotic dexterous operation depends on three core supports: mass-producible tactile perception devices, large-scale physical interaction data infrastructure, and a deployment system covering evaluation and implementation.

Dai Meng has simultaneously unlocked the three high-barrier capabilities, behind which is a global rare tactile intelligence full-stack team spanning robotic operation, tactile perception, large model training, and engineering implementation.

This technical lineage can be traced back to Carnegie Mellon University (CMU).

CMU is the birthplace of global robotics research. In 1979, Turing Award winner and AI pioneer Raj Reddy established the world's first robotics research institute here. Three years later, Matthew T. Mason professor came to serve as the director of the Robotics Research Institute, who is also known as the father of global robotic dexterous operation, and his research laid the theoretical foundation for the entire field of robotic operation, grasping, and contact mechanics.

Mason professor's academic lineage almost supports half of the global robotics industry, from Boston Dynamics, Google DeepMind, Physical Intelligence, NVIDIA, Amazon Robotics and other industry giants, to the touch and operation laboratories of various companies, the entrepreneurs and researchers at the center of today's embodied intelligence stage mostly come from this lineage.

Among Mason's students, a Chinese figure has emerged — Wang Yu, the first doctoral graduate of Mason professor. As a successor to the CMU academic lineage, Wang Yu later brought the entire dexterous operation academic system back to China, becoming the guide of this Chinese tactile entrepreneurship story.

Mason and his two generations of students' academic lineage

Mason's robotics operation school has a core judgment: The dexterity of the hand mainly depends on 'the brain', not the hand itself. This also raises a more fundamental question: if the key to dexterous operation lies in 'the brain', what does this 'brain' learn from?

Language models have books, video models have the internet, but data on contact, force, deformation, and sliding can only be created and accumulated through real operations. Therefore, the first step toward physical intelligence is not to directly create a 'brain', but to first establish a data entry point that can perceive and record real contact.

In 2015, Wang Yu and Li Zexiang co-founded the Hong Kong University of Science and Technology Robotics Research Institute and served as the founding director. It was during this period that Wang Yu began leading the HKUST research team to tackle high-resolution tactile vision perception technology.

In the long-term research efforts that followed, Wang Yu gradually cultivated a young team of scientists covering robot manipulation, multimodal large models, and tactile perception.

In late 2023, Daimeng Robotics officially launched large-scale operations in Shenzhen. Professor Wang Yu served as a co-founder and chief scientist, and together with his three 90s-born doctoral students, formed the core of the team.

Original Weihao, Wang Yu, Duang Jianghua, Du Yipai

Duang Jianghua, a doctoral graduate from the Chinese Academy of Sciences and a postdoctoral researcher at Hong Kong University of Science and Technology, who was selected as a young scientist in the MIT AI100, serves as the founder and CEO, and is the key figure driving the industrialization of tactile vision technology; Original Weihao, who previously served as a multimodal research expert at Alibaba Tongyi Lab and has experience in both large models and physical robot operations, serves as Chief AI Scientist; Du Yipai, a core disciple of Wang Yu in tactile sensing and the chief architect of the monochrome light tactile vision sensor architecture at Daimeng, serves as the R&D and tactile lead.

From Mason's theory of robot manipulation to Wang Yu's tactile technology path, and then to Duang Jianghua, Original Weihao, and Du Yipai respectively driving industrialization, model development, and product engineering, this team spent ten years transforming a scientific proposition about robot dexterity into a fully operational commercial system.

Global First 'Physical Interaction Brain'

Redefining the Scaling Curve of Embodied Intelligence

Continuing to climb along the technical path of 'first establishing a data entry point, then moving towards physical intelligence,' Daimeng recently officially launched the world's first tactile-grounded world model Daimon-TWM (Tactile-grounded World Model), integrating native tactile perception throughout understanding, reasoning, prediction, and verification, thus building the 'tactile nervous system' from fingertip perception to brain reasoning and back to action control, pushing embodied intelligence from 'seeing the world' to 'understanding the world' and 'interacting with the world'.

The model is based on Daimeng's largest-scale multimodal physical world dataset containing tactile data, demonstrating the ability to perform stable operations and generalize in complex scenarios, and showing a continuous scaling trend and cross-scenario, cross-body application potential.

In the task of cleaning broken glass, the transparency and reflection of glass, along with the irregular edges, make it difficult for vision to stably identify the edges of the glass; glass is slippery and fragile, and a light grip can easily cause it to slip, and a slight deviation in gripping force and lifting posture may lead to secondary breakage; the fragments are tightly attached to the tablecloth, and directly lifting them may also pull up the tablecloth along with them.

Daimeng's 'Physical Interaction Brain' Daimon-TWM confirms whether the glass is securely gripped, predicts the consequences of different gripping angles, lifting directions, and placement methods; when it judges that directly lifting may pull up the tablecloth, it actively adjusts the glass to a vertical position before lifting, and continuously refines the force and trajectory based on real-time tactile feedback, finally safely placing the fragments into the trash bin on the desk.

Daimon-TWM autonomously completes the task of cleaning broken glass

Daimon-TWM achieves collaborative interaction through a three-layer architecture, forming a physical intelligence system with physical cognition as the foundation and an operational closed-loop of 'slow planning - fast correction':

Daimon-TWM three-layer architecture

Physical Cognition Module: Based on reinforcement learning, the tactile reasoning foundation is responsible for establishing physical common sense, understanding contact states, force and deformation, material properties, and other key features, forming a cognitive judgment of the current physical state.

Prediction and Decision Module: Based on the world model, the dynamic prediction core is responsible for prospective prediction, predicting the evolution trends of contact states and potential failure risks, and generating action strategies based on the prediction results.

Instantaneous Control Module: Based on tactile feedback, the high-frequency servo system is responsible for immediate action correction, continuously integrating decision instructions with real-time tactile feedback, achieving millisecond-level minor corrections at a frequency of 100Hz.

Daimon-TWM has a parameter scale of 10B level, can perform real-time inference on the NVIDIA RTX 5090 platform, and has the potential for cross-body and cross-robot arm deployment. The model introduces a unified tactile token (unified tactile tokens), extracting contact-related information from multimodal observations, building a unified tactile latent space for tactile understanding and prediction.

Daimon-TWM integrates tactile perception throughout the model, granting the model significant advantages in physical understanding, prediction planning, and operational generalization: Daimon-TWM achieves comprehensive advantages in physical cognition evaluations and demonstrates outstanding autonomous operation capabilities in contact-intensive operational tasks, with an average success rate that is twice as high as π0.5 under disturbance-free conditions and ten times higher under disturbed conditions.

Building a Self-Evolving Loop of Physical Intelligence

In this intense competition led by the 'Physical Interaction Brain,' Daimeng not only builds a first-mover advantage in intelligence but also leads in running the complete loop of 'perception generates data, data refines intelligence, evaluations test capabilities, and deployment promotes iteration,' promoting the 'Cambrian Explosion' of physical intelligence.

Daimon-TWM evolution loop

Based on the already large-scale deployment of tactile perception devices and data collection systems, Daimeng has built the largest-scale multimodal physical world dataset containing tactile data, Daimon-Infinity, covering the complete data pyramid of 'simulation data—no-body data—remote operation data,' and has accumulated tens of thousands of hours of multimodal physical world operation data with tactile content, with plans to expand the data scale to millions of hours within the year, becoming an important foundation for the model to learn real operational patterns and establish interactive capabilities.

Daimon-Infinity's first 10,000 hours of data open-sourced, achieving nearly 50 million downloads on the Alibaba ModelScope community, firmly ranking first in embodied data sets.

In the data collection phase, Daimon also co-built with China Mobile the 'Data Collection into Homes' outbound data collection network, and established the world's first 'Embodied Data Collection 5S Store' in Hunan Chenzhou. The project's first phase plans to deploy 1,000 devices, and after full production, it is expected to generate 1 million hours of real operational data annually.

This means that Daimon is transforming the past data collection, which heavily relied on engineers and fixed sites, into a standardized, flexible production network that can break through time and space limitations.

Daimon also built the industry's first benchmark for physical interaction capabilities with tactile full-modal evaluation, RobOmni, providing a unified standard for model training and optimization.

From perception, data, evaluation to deployment, Daimon has received support from industry partners and collaborators. In the perception phase, Lenovo provides industrial-level mass production capabilities; in the data phase, Daimon co-built an outbound data collection network with China Mobile; in the evaluation phase, NVIDIA provides ecosystem support; in the deployment phase, Huiguang connects industrial automation and precision manufacturing, and China Merchants connects port and logistics scenarios.

From mass production scale to market share, to data and model capabilities, Daimon has achieved 'Global First' in 8 key indicators:

First to achieve mass shipment of visual-tactile perception devices in the tens of thousands;

First to achieve visual-tactile coverage for 80% of dexterous hand manufacturers;

Global first in shipment volume of visual-tactile perception devices;

Global first in revenue scale of the visual-tactile track;

Global first in download volume of embodied data sets with tactile;

Established the world's first large-scale outbound data collection system;

Launched the world's first full-modal tactile benchmark for physical interaction capabilities, RobOmni;

Launched the world's first 'Physical Interaction Brain' Daimon-TWM that supports cross-body and cross-configuration deployment.

With the comprehensive leading advantages in production yield rate, unit cost, and delivery stability, Daimon has served over 200 global clients, with more than 50 overseas clients, and has completed deliveries to global top enterprises and institutions such as OpenAI, Figure, Physical Intelligence, Skild AI, Meta, BMW, and Google DeepMind, while also advancing multiple model deployments and POC collaborations, entering the supply chain of top industrial automation and robotics companies such as Huiguang and Tesla.

Daimon's commercialization curve

This extensive global client network has become a strong proof of Daimon's Physical AI full-stack capabilities and the three-stage leap from 'delivering tactile devices' to 'producing model fuel' and then to 'delivering model capabilities.' Supporting this achievement is Daimon's accelerating replication of standardized commercialization capabilities and its continuous self-evolution of the physical intelligence iteration flywheel.

Perception creates scale effects, data forms network effects, and intelligence opens up value limits. Tactile thus becomes a platform-level compound interest competition among entry points, data, intelligence, and standards.

This article comes from the WeChat official account “Investment Circle” (ID: pedaily2012), author: Yang Ji Yun, authorized by 36Kr for release.

The views in this article represent the author's own, and the 36Kr platform only provides information storage space service.

A good article needs your encouragement

[](https://m.36kr.com/user/10623839)Investment Circle Invited Author

22_Collect_+10_Comment_

_Open WeChat 'Scan', open the web page and click the share button at the top right__Weibo_

_Immersive Reading_ _Return to Top_

Report

Participate in Comments

Comments Millions of Lines, Friendly First

Submit Comment 0/1000

Recent Content

Lei Jun Heavy Investment in a 31-Year-Old Female Executive

5 Hours Ago

Snatching Tsinghua 00s

2026-09-30

Investors Begin to Block Yao Shunyu

2026-09-30

36Kr AI Evaluation

Select Reliable AI, See Real Reviews

View

36Kr AI Evaluation Official Community

Join

36Kr Project Recommendation

Consult Project Review and Entry

Contact

36Kr Project Recommendation Subscription Number

Reported Projects

[Ant Group Contact Global Leading Financial Technology Open Platform](https://pitchhub.36kr.com/project/2144768335758592)

The Big Shake-up in the Mini-Vehicle Market, Mini-Vehicle Sales Plummeted by 55%, Automakers Collectively 'Extended' to Leave the A00 Class Market

The Era of A00 and A-Class Cars Has Come to an End.

2026-08-11

About 36Kr

Report Portal for Internet Rumor Information

Hot Recommendations

Partners

Download 36Kr APP

iOS & Android

This site is provided by Alibaba Cloud for computing and security services. Hotline for reporting illegal and bad information, and protection of minors: 010-89650707. Reporting email: [email protected]Report Online Harmful Information

[](https://weibo.com/wow36kr)[](https://m.36kr.com/rss-center)

Feedback

36Kr APP Let a Part of People See the Future First

36Kr

Jingzhun

Ke Space

Pushing and Interpreting Cutting-Edge, Informative Technology Investment News

Provider of Financial Information and Systems Services for the Primary Market

Focusing on Global Excellent Entrepreneurs, Project Financing Rate Approaches 97%, Leading the Industry