China AI Native Industry Insights – 20250611 – Alibaba | Monica | Zhipu AI | more

Explore Alibaba’s innovative MaskSearch framework for AI-driven active search, Manus’s integration of Google Veo3 for enhanced cinematic AI videos, and Zhipu’s CoCo AI Assistant automating work reports with precision. Discover more in Today’s China AI Native Industry Insights.
1. Alibaba’s MaskSearch Framework: Teaching AI to ‘Find Answers’ Through Active Search
🔑 Key Details:
– New Pretraining Framework: MaskSearch enables AI to learn active searching and multi-step reasoning to answer complex questions accurately.
– Novel RAMP Task: Retrieval-Augmented Mask Prediction teaches models to use search engines to fill in missing information through structured reasoning.
– Multi-Agent System: Combines planner, rewriter, and observer agents to generate clear reasoning paths (Chain-of-Thought).
– Small Model Performance: Even 1.5B parameter models show 11.78% improvement on datasets like Bamboogle after MaskSearch training.
💡 How It Helps:
– AI Researchers: Open-source framework offers a new paradigm for training retrieval-enhanced language models.
– LLM Developers: Curriculum learning strategy provides a structured approach to building search capabilities into smaller models.
– NLP Engineers: DAPO algorithm with hybrid rewards system creates more effective reinforcement learning pipelines.
– Product Teams: Enables smaller, more efficient models to challenge larger ones in open-domain QA tasks.
🌟 Why It Matters:
MaskSearch represents a significant advancement in addressing a fundamental limitation of language models – their inability to access up-to-date information. By teaching AI to actively search rather than hallucinate, this framework bridges the gap between static training data and dynamic information needs. The successful scaling to smaller models demonstrates a path toward more efficient, accurate AI assistants capable of real-world problem-solving without requiring massive parameter counts.
Original Chinese article: https://mp.weixin.qq.com/s/jabNf48U-HoN9W648dmX2Q
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%2FjabNf48U-HoN9W648dmX2Q
Video Credit: The original article
2. Manus Integrates Google Veo3 for Next-Level Cinematic AI Videos
🔑 Key Details:
– Veo3 Integration: Manus now integrates Google’s Veo3 model to enhance video generation quality and realism.
– Cinematic Output: Users can generate longer scenes, synchronized audio, and expressive characters with lifelike detail.
– Membership Access: Available to Basic, Plus, and Pro members—bringing premium cinematic tools to a wider creative base.
– Enhanced Storytelling: Scenes now support richer visual depth and improved narrative pacing for complete storytelling experiences.
💡 How It Helps:
– Creators: Elevates short-form content into immersive narratives with cinematic quality.
– Marketers: Produces more compelling brand storytelling with polished visuals and fluid dialogue.
– Educators: Turns educational material into engaging, high-fidelity visual content.
– Storytellers: Provides an intuitive way to bring scripts and concepts to life without video editing skills.
🌟 Why It Matters:
By integrating Veo3, Manus pushes generative video closer to professional film production quality, making it easier for individuals and teams to tell complex stories. This upgrade underscores the growing convergence between AI tools and creative industries, lowering the barriers to high-end video creation.
Original article: https://x.com/ManusAI_HQ/status/1932091372766720454
Video Credit: ManusAI (@ManusAI_HQ on X)
3. Zhipu’s CoCo AI Assistant: Automating Work Reports with Precision
🔑 Key Details:
– AI Automation: CoCo serves as an AI work assistant that automatically compiles comprehensive work reports.
– Multiple Formats: CoCo generates reports in various formats including text summaries, data analyses, and visual presentations.
– Customizable Templates: The system offers customizable report templates tailored to different business scenarios and reporting needs.
– Data Analytics: CoCo includes capabilities for analyzing performance metrics and providing optimization suggestions based on data.
💡 How It Helps:
– Team Managers: Reduces administrative overhead by automating routine reporting tasks while maintaining professional quality.
– Data Analysts: Provides streamlined data visualization and analysis, saving hours of manual compilation work.
– Project Coordinators: Enables quick generation of status updates and progress reports with consistent formatting.
– Business Executives: Delivers clear, concise summaries of complex information for efficient decision-making.
🌟 Why It Matters:
As organizations seek to optimize productivity, AI assistants like CoCo represent a significant shift in workplace efficiency. By automating the time-consuming process of report generation, companies can redirect valuable human resources toward strategic initiatives rather than administrative documentation. This technology particularly addresses the pain point of report standardization across teams while adapting to specific business contexts.
Original Chinese article: https://mp.weixin.qq.com/s/3COowx-TJcJ2y13RvsXX2g
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%2F3COowx-TJcJ2y13RvsXX2g
Video Credit: Zhipu Wechat Channel
4. SiliconFlow Secures Hundreds of Millions in Series A Funding Led by Alibaba Cloud
🔑 Key Details:
– Investment: SiliconFlow completed Series A funding of hundreds of millions RMB led by Alibaba Cloud with participation from existing investor Innovation Works.
– User Growth: SiliconCloud platform has achieved 6 million users and thousands of enterprise clients, with daily token generation exceeding 100 billion.
– Technology Focus: Company developed high-performance inference engines optimized for domestic chips, transforming Chinese computing power from “usable” to “excellent”.
– Product Portfolio: Offers SiliconCloud (cloud.siliconflow.cn) with 100+ mainstream models and BizyAir (bizyair.cn) for AI design workflows.
💡 How It Helps:
– AI Developers: One-stop platform for model fine-tuning, hosting, and deployment, significantly reducing development barriers.
– Enterprise Clients: Multiple deployment options including API services, dedicated instances, software subscriptions, and integrated LLM hardware solutions.
– Content Creators: Cloud GPU resources that integrate with local ComfyUI, solving local computing limitations for AI designers.
– IT Infrastructure Teams: Heterogeneous computing management platform with elastic scheduling for better resource utilization.
🌟 Why It Matters:
SiliconFlow’s advances in domestic chip optimization and cloud-native AI tools position it as a key player in China’s evolving AI stack. Its ability to unify fragmented resources into scalable infrastructure bridges the gap between cutting-edge AI and real-world applications—fueling wider access, faster iteration, and deeper integration across industries.
Original Chinese article: https://mp.weixin.qq.com/s/zBTNc9sHoEAnJLvHU1r8Vg
English translation via free online service: https://translate.google.com/translate?hl=en&sl=zh-CN&tl=en&u=https%3A%2F%2Fmp.weixin.qq.com%2Fs%2FzBTNc9sHoEAnJLvHU1r8Vg
Video Credit: The original article
That’s all for today’s China AI Native Industry Insights. Join us at AI Native Foundation Membership Dashboard for the latest insights on AI Native, or follow our linkedin account at AI Native Foundation and our twitter account at AINativeF.