On August 24, 2026, People’s Daily Overseas Edition released an AI-themed documentary recording Professor Carles Sierra’s in-depth field study of real-world AI education applications in China. Professor Sierra is also Director of the Artificial Intelligence Research Institute in Barcelona, and has previously served as a council member and program chair of the International Joint Conference on Artificial Intelligence (IJCAI), one of the most prestigious academic conferences in the field of AI.
In addition, as a member of the IEEE working group on standards for large AI models in education, Professor Carles visited Shantou Keli Experimental School, an “AI Education 5A-Level Experimental Base” co-established with Squirrel Ai. The school is a representative case of Squirrel Ai’s large-scale application in China’s K-12 education system. He also visited Squirrel Ai’s intelligent self-study center in Tonglu, Hangzhou, where he closely observed the application of Squirrel Ai’s adaptive learning system in a self-directed learning environment, and engaged in discussions with students and supervising teachers. He gave high recognition to Squirrel Ai’s L5-level autonomous learning model and its outcomes.
“I have been researching artificial intelligence for the past 40 years, with a particular focus on its applications in education,” Professor Carles said in the documentary. “I came to China to see how AI is being applied in real educational practice.”
In the classroom at Keli Experimental School, Professor Carles observed a deeply integrated AI-powered education system embedded into daily teaching routines. After students complete diagnostic assessments on smart learning devices, Squirrel Ai’s adaptive system analyzes multimodal data, including response patterns, error types, and time spent per question, to infer each student’s knowledge mastery and precisely identify learning gaps. In doing so, Squirrel Ai is translating China’s centuries-old educational ideal of “teaching students in accordance with their aptitude” into a tangible model of future education.
During the field visit, a teacher demonstrated a scenario to Professor Carles Sierra: in traditional teaching, when a student’s answer is incorrect, teachers typically only mark it as right or wrong. However, Squirrel Ai’s educational logic goes far beyond binary evaluation. It combines each student’s learning profile to conduct personalized diagnosis, trace the root cause of the mistake, and identify the underlying knowledge gap. It then automatically delivers targeted explanations and practice exercises to help students address the issue at its source, preventing repeated errors of the same type.
In the discussion, teachers from Keli Experimental School also noted that all content recommendations generated by Squirrel Ai are fully driven by algorithms, requiring no manual intervention. From their perspective, AI serves as an assistive teaching tool. The core value of Squirrel Ai lies in providing teachers with precise learning analytics, enabling them to focus their energy on emotional engagement with students and guiding higher-order thinking skills. They emphasized that this level of autonomy does not diminish the role of teachers; rather, it redefines it.
Squirrel Ai’s Technological Depth
During the visit, Professor Carles Sierra held in-depth discussions with Squirrel Ai Chairman Li Haoyang. Li noted that Squirrel Ai’s multimodal adaptive education large model is built on a fine-grained knowledge graph architecture, which decomposes subject knowledge into micro-level cognitive units. This enables the system to construct a personalized knowledge profile for each student. Combined with a question bank of over 500 million items, designed with the participation of outstanding teachers and powered by AI systems capable of addressing massive individualized learning demands, the model is able to generate precisely tailored learning pathways for each student in large-scale educational settings, making truly personalized education possible.
Notably, in 2024, the IEEE Working Group on Standards for AI Large Models in Education was officially established, with Squirrel Ai founder Li Haoyang serving as its chair. The working group includes leading figures such as former Carnegie Mellon University School of Computer Science Dean Tom Mitchell, former President of the Association for the Advancement of Artificial Intelligence (AAAI) Bart Selman, President of the European Association for Artificial Intelligence Carles Sierra, Harvard Graduate School of Education Professor Chris Dede, former President of SRI International Stephen Ciesinski, as well as chief scientists and chief learning officers from leading global education organizations including Pearson, ETS, and Khan Academy.
This marks the first time a Chinese company has taken a leading role in shaping international standards for AI in education. It positions Squirrel Ai not only as a technology practitioner, but also as a co-architect of global industry standards.
A Model for Scaled Education
The AI-powered education scenario at Keli Experimental School represents one of Squirrel Ai’s large-scale validation pathways. To date, Squirrel Ai has served more than 43 million users, accumulated over 20 billion learning behavior data points, established more than 3,000 intelligent learning centers nationwide, and formed in-depth collaborations with over 60,000 public schools. This extensive user base and massive data accumulation enable Squirrel Ai’s multimodal adaptive education large model to continuously iterate in real-world teaching environments.
At Wenyuan High School in Liaocheng, Shandong Province, an experimental group of 84 students used Squirrel Ai’s multimodal adaptive education large model during two weekly math self-study sessions, while a control group of 198 students followed traditional instruction. Midterm exam results showed that the experimental group achieved an average score of 104.7 in the content they learned with Squirrel Ai, compared to 73.55 in the content not covered by Squirrel Ai, representing a 31.2-point difference (out of 120). These findings were published in the academic journal China Educational Informatization.
At Baishaping Primary School in Badong County, Hubei Province, Squirrel Ai introduced its multimodal adaptive education large model to provide targeted instruction for sixth-grade students over a one-month period. According to experimental results from CCTV’s program Super Brain, the students’ average math score increased from 56 to 89. These two cases demonstrate that across different conditions and educational stages, Squirrel Ai has delivered outcomes that are both verifiable and practically implementable.
As a representative of Chinese AI education companies entering the global stage of education standard-setting, Squirrel Ai’s exploration goes far beyond technology itself. It provides a replicable model for the digital transformation of basic education in China and contributes Chinese experience and solutions to the global development of AI in education.
On the path of empowering education through technology, Squirrel Ai has consistently adhered to its core mission: “help every child become a better version of themselves.” It ensures that technology serves human development, that intelligence safeguards the warmth of education, and that it leaves a meaningful imprint on the future shape of learning.
