MiningLamp Technology has added a major credential to its fast-growing reputation in enterprise AI. The Hong Kong–listed company (2718.HK) took home the Grand Prize at the national finals of the 3rd China’s Innovation Challenge on Artificial Intelligence Application Scene (CICAS)—one of the country’s most competitive and influential AI events—cementing its status as a serious force in Agentic AI and multimodal large models.

The winning project, developed in collaboration with Peking University, is called “Intelligent Platform for Brand Globalization Creative Generation and Emotional Connection Based on Multimodal Large Models.” Beyond the long name, the idea is straightforward and timely: help companies expand globally by using AI to localize creative content, predict emotional response, and generate marketing assets faster—without losing cultural nuance.

The project was also named a “2025 National Artificial Intelligence Application Scenario Exemplary Case,” a designation reserved for AI systems with strong real-world commercial and societal impact.

A Fierce Competition With National Significance

CICAS is not a typical startup pitch contest. Jointly organized by the Chinese Association for Artificial Intelligence, the Suzhou Municipal People’s Government, and Soochow University, the competition is designed to surface AI technologies that can scale across industries.

This year’s challenge drew more than 3,250 registered teams, with 113 elite teams advancing to the national finals. Over 350 participants—from China and abroad—competed in Suzhou, Jiangsu Province, placing MiningLamp’s win firmly in “best-of-the-best” territory.

For MiningLamp, this marks a symbolic moment. While the company has been active in enterprise AI since 2006, the CICAS Grand Prize represents its first major national AI competition win since its Hong Kong Stock Exchange listing in November 2025.

Why Brand Globalization Is an AI Problem Now

Global expansion has become harder, not easier, for brands. Cultural missteps go viral instantly, consumer sentiment shifts faster than traditional research can track, and content localization remains expensive and slow.

MiningLamp’s platform is built around a clear thesis: global brand marketing is no longer a creative-only problem—it’s a data, emotion, and automation problem.

According to Wu Minghui, Founder, CEO, and CTO of MiningLamp, brands going global face three persistent barriers:

• Cultural and emotional differences across markets

• High costs and long timelines for content localization

• Limited data-driven insight into how creative will actually land

The platform addresses these challenges by combining multimodal AI, Agentic workflows, and proprietary data intelligence into a single system designed for marketing teams—not just data scientists.

Four Technologies Powering the Platform

At the core of MiningLamp’s winning solution are four tightly integrated capabilities. Together, they form an end-to-end workflow that spans insight generation, emotional evaluation, and content creation.

1. A Global Multimodal Content Asset Library

The platform includes a multimodal content library covering major global markets, incorporating video, image, and text assets. Rather than starting from scratch for every campaign, brands can draw from culturally relevant creative materials aligned with regional norms and preferences.

The practical upside is speed and cost efficiency. Localization cycles that once took weeks can now be compressed into days—or even hours—while maintaining cultural relevance.

In a market where brands are expected to “think global but act local,” this library becomes a strategic advantage rather than a simple repository.

2. Trustworthy Data Collection via the “Dexterous Hand” Mano Model

MiningLamp’s Mano model—described internally as an AI “dexterous hand”—is one of the platform’s most distinctive features.

Mano can operate across browser environments, visually identifying interface elements and interacting with them much like a human would. Users simply provide a URL and a description of their data needs; Mano handles the rest, collecting multi-source web data with minimal manual intervention.

This capability matters because global market analysis often fails due to fragmented, unreliable data. Mano’s human-like perception allows it to gather cleaner, more contextual datasets—critical for downstream decision-making.

Technical benchmarks underscore Mano’s maturity:

• Ranked first in the specialized model category of the OSWorld benchmark

• Ranked second overall, just behind Anthropic’s Claude-Sonnet-4.5

• Achieved SOTA performance on the Mind2Web benchmark

• A 7B-parameter version supports private deployment for enterprise security needs

For enterprises wary of black-box AI, this emphasis on transparency and controllability is notable.

3. Subjective Emotional Perception Analysis With HMLLM

Perhaps the most ambitious element of the platform is its Hypergraph Multimodal Large Language Model (HMLLM), designed to simulate how different audiences feel about content—not just how they engage with it.

Unlike traditional sentiment analysis, HMLLM models subjective emotional response across dimensions like attention, emotion, and cognition. It can estimate how viewers from different cultures, age groups, and genders are likely to react to advertising content before it goes live.

The model is trained on uniquely rich datasets:

• Video-SME and SPA-ADV, built from EEG and eye-tracking data

• Data collected from over 10,000 real human subjects

• Emotional response modeling with R² consistency exceeding 89%

The research behind HMLLM earned a Best Paper Nomination at ACM MM 2024, lending academic credibility to what is often treated as a fuzzy marketing problem.

For global brands, the implication is clear: fewer cultural misfires, less guesswork, and more confidence in creative decisions.

4. AI-Driven Video Generation at Scale

Once insights and emotional assessments are complete, the platform can automatically generate and optimize video content. This closes the loop—from market understanding to creative output—inside a single AI-driven workflow.

MiningLamp claims the system can compress traditional video production timelines from weeks to hours, a meaningful advantage as short-form video and rapid campaign iteration become standard across platforms like TikTok, YouTube, and connected TV.

This positions the platform not just as an analytics tool, but as a full-stack AI marketing engine.

Enterprise-Grade AI, Not Just a Demo

MiningLamp’s broader technical credentials reinforce the seriousness of the platform. The company has published 20+ papers in top-tier international journals and conferences, including:

• ACM MM 2024 (CCF-A): Best Paper Nomination for HMLLM

• TPAMI (SCI Q1): Few-shot video instance segmentation

• IJCV (SCI Q1): Image generation methods

• AAAI 2026 (CCF-A): Mano model compression, accepted as an oral presentation

These aren’t marketing whitepapers—they’re peer-reviewed contributions that help explain why MiningLamp is increasingly described as China’s first “Agentic AI” public company.

Strategic Implications for MarTech and Global Brands

MiningLamp’s win highlights a broader industry trend: AI is moving from content optimization to content decision-making.

While many Western MarTech platforms focus on performance metrics after launch, MiningLamp is betting on AI that evaluates cultural fit and emotional resonance before content reaches consumers. That shift could reshape how global campaigns are planned, especially in regulated or reputation-sensitive industries.

The platform also aligns with the growing enterprise demand for trustworthy AI—systems that are explainable, auditable, and deployable in private environments.

What Comes Next

At the CICAS closing ceremony, MiningLamp signed a cooperation intent with Gusu District, signaling plans to expand AI R&D and real-world deployment scenarios locally.

The company says it will continue applying its “data-driven trustworthy productivity” philosophy beyond brand globalization, targeting additional vertical industries where Agentic AI can deliver measurable impact.

As global competition intensifies and AI-driven differentiation becomes table stakes, MiningLamp’s platform could emerge as a critical infrastructure layer for companies trying to scale internationally without losing cultural intelligence along the way.

Get in touch with our MarTech Experts.