๐ŸผStalecollected in 20m

Big Tech AI Battle for China's Gaokao Season

Big Tech AI Battle for China's Gaokao Season
PostLinkedIn
๐ŸผRead original on Pandaily

๐Ÿ’กSee how top Chinese AI labs are optimizing LLMs for high-accuracy educational tasks during peak demand.

โšก 30-Second TL;DR

What Changed

Major Chinese tech firms are deploying AI for Gaokao exam prep

Why It Matters

The integration of LLMs into high-stakes education creates a massive real-world testing ground for AI reasoning and accuracy, influencing future consumer AI adoption.

What To Do Next

Analyze how these companies handle hallucination in educational content to improve your own RAG pipeline reliability.

Who should care:Developers & AI Engineers

Key Points

  • โ€ขMajor Chinese tech firms are deploying AI for Gaokao exam prep
  • โ€ขFocus on personalized tutoring and automated study assistance
  • โ€ขHigh-stakes competition for user acquisition during exam season

๐Ÿง  Deep Insight

Web-grounded analysis with 20 cited sources.

๐Ÿ”‘ Enhanced Key Takeaways

  • โ€ขChinese tech firms are implementing safety restrictions on their AI platforms, such as disabling photo-based question answering, during the Gaokao exam period to prevent cheating and ensure fairness.
  • โ€ขBeyond exam preparation, AI is increasingly utilized for post-Gaokao services, including personalized university and major application assistance, leveraging student scores and regional admission data.
  • โ€ขThe China AI in education technology market is experiencing significant growth, valued at USD 846.1 million in 2025 and projected to reach USD 16,252.10 million by 2034, with a compound annual growth rate of 38.87%.
  • โ€ขAI is being integrated into Gaokao security and monitoring, with AI-assisted invigilation systems deployed in exam sites to detect suspicious behaviors like unusual eye movements or prohibited device usage.
  • โ€ขAI tools have evolved from simple Q&A bots to more sophisticated multi-turn, agent-style deep search systems capable of maintaining context and conducting iterative research for comprehensive support.

๐Ÿ› ๏ธ Technical Deep Dive

  • AI tools have evolved from basic single-round Q&A bots to advanced multi-turn Agent-style deep search systems that can maintain context across extended conversations, ask clarifying questions, and perform iterative research.
  • Alibaba's multilingual large language model, Qwen2-72B, demonstrated strong performance in a 2024 Gaokao simulation for Chinese literature and English language, achieving an average accuracy rate of about 72%.
  • Baidu's Wenxin 4.0 also performed well in a 2024 Gaokao simulation, notably scoring highest in mathematics, physics, and biology among tested AI models.
  • AI-assisted invigilation systems deployed in Gaokao exam sites utilize machine-learning algorithms to analyze video feeds in real-time, identifying suspicious behaviors such as unusual eye movements, handwriting inconsistencies, or the use of forbidden electronic devices.
  • Adaptive learning systems, such as those developed by Squirrel AI, break down subjects into thousands of granular 'knowledge points' to precisely diagnose student understanding gaps and dynamically tailor personalized lesson plans.

๐Ÿ”ฎ Future ImplicationsAI analysis grounded in cited sources

The integration of AI into Gaokao preparation will lead to increased educational equity.
AI tools can provide personalized learning and university application guidance, potentially bridging information gaps for students in rural or less developed areas with limited access to professional counseling.
Regulatory bodies will continue to strengthen measures to prevent AI-assisted cheating in high-stakes exams.
Given the current implementation of restrictions on AI features during Gaokao and warnings from the Ministry of Education, authorities are likely to evolve their anti-cheating strategies as AI technology advances.
Chinese AI models will continue to improve their performance in complex reasoning tasks, particularly in subjects like mathematics.
While AI models currently struggle with complex math and classical Chinese, ongoing research and competition among tech giants, as seen in Gaokao simulations, will drive advancements in these areas.

โณ Timeline

2015-12
iFLYTEK announces goal for AI to pass Gaokao and enter top universities by 2020, focusing on cognitive intelligence.
2017-06
AI-MATHS, an AI program, publicly takes the math section of the Gaokao, scoring 105 out of 150 for Beijing's version of the exam.
2020-07
Baidu launches an AI-powered service to assist Gaokao candidates in choosing universities and majors.
2021-01
Yuanfudao, an AI-based education service, offers free consultations and livestreams to support Gaokao students.
2024-06
Alibaba's Qwen2-72B and Baidu's Wenxin 4.0 participate in a Gaokao simulation, demonstrating strong performance in language but challenges in mathematics.
2025-06
Major Chinese AI platforms (Tencent's Yuanbao, Alibaba's Qwen, ByteDance's Doubao, DeepSeek, Kimi) temporarily disable photo-recognition and question-answering features during Gaokao to prevent cheating.
๐Ÿ“ฐ

Weekly AI Recap

Read this week's curated digest of top AI events โ†’

๐Ÿ‘‰Related Updates

AI-curated news aggregator. All content rights belong to original publishers.
Original source: Pandaily โ†—