Connect with us

Media OutReach

PolyU develops novel multi-modal agent to facilitate long video understanding by AI, accelerating development of generative AI-assisted video analysis

Published

on

HONG KONG SAR – Media OutReach Newswire – 10 June 2025 – While Artificial Intelligence (AI) technology is evolving rapidly, AI models still struggle with understanding long videos. A research team from The Hong Kong Polytechnic University (PolyU) has developed a novel video-language agent, VideoMind, that enables AI models to perform long video reasoning and question-answering tasks by emulating humans’ way of thinking. The VideoMind framework incorporates an innovative Chain-of-Low-Rank Adaptation (LoRA) strategy to reduce the demand for computational resources and power, advancing the application of generative AI in video analysis. The findings have been submitted to the world-leading AI conferences.

A research team led by Prof. Changwen Chen, Interim Dean of the PolyU Faculty of Computer and Mathematical Sciences and Chair Professor of Visual Computing, has developed a novel video-language agent VideoMind that allows AI models to perform long video reasoning and question-answering tasks by emulating humans’ way of thinking. The VideoMind framework incorporates an innovative Chain-of-LoRA strategy to reduce the demand for computational resources and power, advancing the application of generative AI in video analysis.

Videos, especially those longer than 15 minutes, carry information that unfolds over time, such as the sequence of events, causality, coherence and scene transitions. To understand the video content, AI models therefore need not only to identify the objects present, but also take into account how they change throughout the video. As visuals in videos occupy a large number of tokens, video understanding requires vast amounts of computing capacity and memory, making it difficult for AI models to process long videos.

Prof. Changwen CHEN, Interim Dean of the PolyU Faculty of Computer and Mathematical Sciences and Chair Professor of Visual Computing, and his team have achieved a breakthrough in research on long video reasoning by AI. In designing VideoMind, they made reference to a human-like process of video understanding, and introduced a role-based workflow. The four roles included in the framework are: the Planner, to coordinate all other roles for each query; the Grounder, to localise and retrieve relevant moments; the Verifier, to validate the information accuracy of the retrieved moments and select the most reliable one; and the Answerer, to generate the query-aware answer. This progressive approach to video understanding helps address the challenge of temporal-grounded reasoning that most AI models face.

Another core innovation of the VideoMind framework lies in its adoption of a Chain-of-LoRA strategy. LoRA is a finetuning technique emerged in recent years. It adapts AI models for specific uses without performing full-parameter retraining. The innovative chain-of-LoRA strategy pioneered by the team involves applying four lightweight LoRA adapters in a unified model, each of which is designed for calling a specific role. With this strategy, the model can dynamically activate role-specific LoRA adapters during inference via self-calling to seamlessly switch among these roles, eliminating the need and cost of deploying multiple models while enhancing the efficiency and flexibility of the single model.

VideoMind is open source on GitHub and Huggingface. Details of the experiments conducted to evaluate its effectiveness in temporal-grounded video understanding across 14 diverse benchmarks are also available. Comparing VideoMind with some state-of-the-art AI models, including GPT-4o and Gemini 1.5 Pro, the researchers found that the grounding accuracy of VideoMind outperformed all competitors in challenging tasks involving videos with an average duration of 27 minutes. Notably, the team included two versions of VideoMind in the experiments: one with a smaller, 2 billion (2B) parameter model, and another with a bigger, 7 billion (7B) parameter model. The results showed that, even at the 2B size, VideoMind still yielded performance comparable with many of the other 7B size models.

Prof. Chen said, “Humans switch among different thinking modes when understanding videos: breaking down tasks, identifying relevant moments, revisiting these to confirm details and synthesising their observations into coherent answers. The process is very efficient with the human brain using only about 25 watts of power, which is about a million times lower than that of a supercomputer with equivalent computing power. Inspired by this, we designed the role-based workflow that allows AI to understand videos like human, while leveraging the chain-of-LoRA strategy to minimise the need for computing power and memory in this process.”

AI is at the core of global technological development. The advancement of AI models is however constrained by insufficient computing power and excessive power consumption. Built upon a unified, open-source model Qwen2-VL and augmented with additional optimisation tools, the VideoMind framework has lowered the technological cost and the threshold for deployment, offering a feasible solution to the bottleneck of reducing power consumption in AI models.

Prof. Chen added, “VideoMind not only overcomes the performance limitations of AI models in video processing, but also serves as a modular, scalable and interpretable multimodal reasoning framework. We envision that it will expand the application of generative AI to various areas, such as intelligent surveillance, sports and entertainment video analysis, video search engines and more.”

Hashtag: #PolyU #AI #LLMs #VideoAnalysis #IntelligentSurveillance #VideoSearch

The issuer is solely responsible for the content of this announcement.

Advertisement

Media OutReach

Southco’s Blind Mate Floating Mechanism Empowers Efficient Cooling in Data Centers

Published

on

HONG KONG SAR – Media OutReach Newswire – 20 January 2026 – Global computing power is growing aggressively and pushing high-density chip power consumption. As this power density increases, traditional air cooling is reaching its physical limits, and once-fringe liquid cooling technology, boasting thermal efficiency hundreds of times greater than air cooling, is rapidly becoming core infrastructure.

Southco’s Floating Mechanism

Driven by global sustainability targets and the need for high-performance, energy-efficient infrastructures, new large data centers must achieve their targeted PUE values, accelerating the industry towards large-scale liquid cooling. However, reliability issues stemming from insufficient mechanical tolerances at the connection points of liquid cooling systems are becoming a critical bottleneck for energy efficiency upgrades and stable operation. Southco recognizes the severity of this challenge and is committed to providing breakthrough solutions.

Minor Deviations, Major Costs

During the large-scale deployment of liquid cooling technology, the reliability of connection interfaces is vital. According to key data from the Open Compute Project (OCP) “Rack-Mounted Manifold Requirements and Verification Guidelines,” a mere 1mm increase in mechanical deviation at liquid cooling interfaces can significantly raise system flow resistance by 15%, leading to a 7% increase in pump energy consumption! This is no trivial amount; in a hyperscale data center with thousands of interfaces, it translates to millions of kilowatt-hours of additional energy consumption and substantial operational costs each year. More concerning is that traditional rigid connection solutions typically offer only ±0.5mm of static tolerance, which proves inadequate in complex real-world environments like these:

  • Accumulation of Multi-Dimensional Installation Deviations: In mixed deployment scenarios of widely used EIA-310-D standard racks and advanced ORV3 open architectures, rack installation tolerances can accumulate up to ±3.2mm, far exceeding the limits of traditional solutions.
  • Dynamic Vibration Impacts: In ISTA 3-E vibration tests simulating real transportation and operating environments, interface displacement often exceeds 2.8mm, posing significant risks of leaks or connection failures.
  • Material Thermal Expansion Effects: Under a typical temperature change of 55°C, copper alloy manifolds can expand approximately 1.2mm per meter, continuously challenging fixed interfaces.

These dynamic, multi-dimensional deviations underline the urgent need for an intelligent, reliable sealing connection solution to ensure the long-term, efficient, and safe operation of liquid cooling systems.

Blind Mate Quick Disconnect: A Connection for a Dynamic World

As a century-old expert in precision engineering, Southco confronts this challenge head-on with the launch of the revolutionary new “Blind Mate Floating Mechanism” liquid cooling connection solution. This innovation is not just a new product; it is a systematic response to profound insight into industry pain points.

Current Status and Trends of Blind Mate Floating Technology

Liquid cooling technology is rapidly gaining traction in high-performance computing (HPC), AI training clusters, and hyperscale data centers. Blind mate technology allows devices to connect without precise visual alignment, making it a core interface solution for rapid deployment and efficient maintenance in liquid cooling systems (especially cold plate systems). The development trends are clear:

  • Higher Tolerance Capacity: Adapting to more complex rack environments and dynamic changes is essential.
  • Increased Reliability: Zero leakage, long lifespan, and resistance to extreme conditions are basic requirements.
  • Intelligent Integration: Integrating sensors for flow, temperature, pressure, etc., for real-time monitoring and predictive maintenance is a future direction.
  • Standardization and Compatibility: Supporting OCP ORV3 and other open standards for seamless integration across platforms and manufacturers.
  • Lightweight and Compact Design: Meeting the demands of higher density deployments.

Southco’s Blind Mate Floating Mechanism exemplifies innovative design centered around these core trends.

Advantages Over Traditional Solutions

Compared to traditional fixed or simple floating heat transfer connection solutions, Southco’s “Blind Mate Floating Mechanism” offers a qualitative leap with advantages evident across multiple dimensions.

  • Three-Dimensional Dynamic Tolerance Control: Southco’s innovative design features a groundbreaking three-dimensional dynamic compensation mechanism: ±4mm of floating tolerance in the radial direction (2° tilt compensation) and 6mm of displacement absorption capacity in the axial direction. This far exceeds common static tolerances in the industry, effectively absorbing and adapting to all previously mentioned installation tolerances, vibration displacements, and thermal expansion deformations.
  • Self-Centering Mechanism: When the liquid-cooled blind-plug connector is unplugged, the floating structure automatically resets to the center position, ensuring sufficient floating space for plugging and unplugging operations, fully meeting the strict tolerance requirements of OCP and ORV3 standards.
  • Outstanding Sealing Performance for Long-Term Operation: Products endure rigorous ASME B31.3 standard 300psig burst pressure tests, ensuring over 10 years of service life, providing long-term stability for data centers, an achievement traditional solutions cannot match.
  • Efficient Maintenance and Significant Cost Reduction: Featuring the Universal Quick Disconnect Blind (UQDB) interface, the design enables genuine “blind operations,” allowing server maintenance without precise alignment or specialized tools. Quick disconnect capabilities make server replacement or upgrades as convenient as “hot plugging,” reducing downtime and related losses by over 90%.

The Necessity of Blind Mate Floating Mechanism

In AI computing clusters and next-generation data centers, adopting advanced connection solutions like Southco’s “Blind Mate Floating Mechanism” is no longer optional, but essential.

  • Key to Overcoming Cooling Bottlenecks: High power density is an inevitable trend; traditional cooling and rigid connections can no longer meet the demand. The Blind Mate Floating Mechanism is foundational for unleashing the full potential of liquid cooling and ensuring efficient, stable operation of cooling systems.
  • Cornerstone of Business Continuity: The costs of data center downtime are immense. The rapid, reliable thermal maintenance supported by the Blind Mate Floating Mechanism is vital for ensuring uninterrupted operation of critical business activities 24/7.
  • Core to Achieving Green and Low-Carbon Goals: Minor deviations in connectors leading to increased flow resistance can significantly elevate pump energy consumption. Its high-tolerance, low-flow resistance design directly contributes to lowering data center PUE values, making it an important element in energy conservation and emissions reduction goals.
  • Flexibility for Future Expansion and Upgrading: Modular and standardized design enables data centers to expand capacity and upgrade equipment more flexibly, easily accommodating future increases in computing power demand and technological iterations.

Continuous Innovation, Intelligent Cooling Ahead

According to the “Open Rack V3” white paper, liquid cooling penetration in hyperscale data centers is expected to exceed 40% by 2025. Southco continues to invest in R&D to iterate floating mechanism technology:

* Exploring lightweight high-performance materials (like PPSU thermoplastic) to reduce weight while maintaining strength.
* Advancing intelligent sensor integration for real-time monitoring of key parameters like flow and temperature, providing data for predictive maintenance and energy efficiency optimization.
* Deepening ecosystem collaboration and standardization to promote liquid cooling interfaces in higher density, lower TCO, and broader applications.

As liquid cooling technology transitions from optional to essential, Southco’s “Blind Mate ” represents a precision engineering product innovation, and a profound understanding of cooling challenges in the data center industry. By integrating over a century of precision mechanical design expertise with innovative three-dimensional dynamic tolerance control technology, Southco strives to help global data centers break through cooling bottlenecks, building a more efficient, reliable, and greener foundation for computing power, empowering infinite possibilities in the digital age.

For more information, please click here for more details.
Hashtag: #southco #liquidcooling #EnterpriseComputing #innovations

The issuer is solely responsible for the content of this announcement.

About Southco

Southco, Inc. is the leading global designer and manufacturer of engineered access solutions. From quality and performance to aesthetics and ergonomics, we understand that first impressions are lasting impressions in product design. For over 75 years, Southco has helped the world’s most recognized brands create value for their customers with innovative access solutions designed to enhance the touch points of their products in transportation and industrial applications, medical equipment, data centers and more. With unrivalled engineering resources, innovative products and a dedicated global team, Southco delivers the broadest portfolio of premium access solutions available to equipment designers throughout the world.

Southco Asia Limited
2401, Tower 2, Ever Gain Plaza
88 Container Port Road, Kwai Chung
Hong Kong

Continue Reading

Media OutReach

Personalizing The Hong Kong Stay Experience With Dorsett Kai Tak’s ‘3 Wishes’ Package

Published

on

Save up to 25% with THREE self-chosen perks to match each journey— whether for family fun or business needs

HONG KONG SAR – Media OutReach Newswire – 20 January 2026 – Dorsett Kai Tak, Hong Kong invites travellers to personalize their Hong Kong stay with the Dorsett 3 Wishes Room Package. Guests can enjoy up to 25% savings on a twonight stay while choosing THREE perks from a list of nine, tailored to their travel purpose—whether for leisure or business—combining modern luxury with authentic local experiences.

Stays Tailored for Every Kind of Traveller

For families and leisure travellers: enjoy local snacks vouchers from Kowloon City’s “Little Thailand”—with treats like egg waffles, pandan buns, or mango‑longan‑coconut dessert—plus options such as complimentary breakfast, room upgrade or a free Octopus card for easy city travel.

For business travellers: choose perks like an Airport Express ticket to Kowloon Station, complimentary suit pressing, or a local data SIM card to stay connected.

Dorsett Kai Tak sits beside the iconic Kai Tak Sports Park, home to world-class concerts and international sporting events. Guests enjoy exclusive walkway access to the Main Stadium and Kai Tak malls. Complimentary shuttle service to AIRSIDE mega mall and Kai Tak MTR ensures easy shopping and city exploration.

Leisure travellers can explore Kowloon Walled City Park for a glimpse of the city’s past or visit the former Kai Tak Airport site for aviation history. Foodies may wander through Kowloon City for Southeast Asian flavours or enjoy traditional Chiu Chow cuisine.

Featuring 373 contemporary rooms, including family-friendly triples, quads, connecting rooms, and balcony suites — ideal for groups of friends or families. Dining options include Jin Bo Law Skybar, Kai Fuk Lau, and Siete Ocho.

Book direct on the official Dorsett Kai Tak website to enjoy a personalized and authentic Hong Kong journey.

Hashtag: #DorsettKaiTak

The issuer is solely responsible for the content of this announcement.

About Dorsett Kai Tak, Hong Kong

This flagship property of Dorsett Hospitality International offers a modern five‑star stay in Hong Kong’s historic Kai Tak district. Conveniently located near Kai Tak Sports Park, the hotel provides exclusive stadium access and a complimentary shuttle to AIRSIDE and Kai Tak MTR.

Guests can choose from 373 stylish rooms and suites—including family-friendly options and the Presidential Pool Suite—and enjoy an infinity pool, 24-hour gym, Jin Bo Law Skybar, and dining at Kai Fuk Lau and Siete Ocho.

Continue Reading

Media OutReach

SNP Supervisory Board Extends Contract of CEO Jens Amail Ahead of Schedule

Published

on

  • Contract of Jens Amail extended until December 2030
  • Early renewal reflects strong operational performance and successful strategic development
  • Partnership with Carlyle marks next phase of SNP’s growth trajectory

HEIDELBERG, GERMANY – Media OutReach Newswire – 20 January 2026 – The Supervisory Board of SNP Schneider-Neureither & Partner SE has extended the contract of Chief Executive Officer Jens Amail ahead of schedule by further five years. Amail, who has been leading the company since January 2023, will now remain CEO until December 2030. The early contract extension reflects the company’s successful development in recent years and signals stability and confidence in its leadership.

Jens Amail, CEO of SNP SE

Willi Westenberger, Managing Director at Carlyle and Chairman of SNP’s Supervisory Board, says: “Under Jens Amail’s guidance, SNP has made significant progress across its key strategic priorities, strengthened its market position and delivered sustained operational improvements. The company has accelerated its international expansion and broadened its global partner ecosystem as the foundation for sustainable growth. Extending his contract is a clear signal of continuity and underlines our confidence in his leadership and the strong team he has built at SNP.”

The significant operational progress and stability gained by SNP are closely linked to the leadership of Jens Amail. Since the end of 2022, the company has achieved annual records in order entry, revenue and EBIT. The sustained increase in profitability was backed up by a successful internationalization and growth strategy. This has opened up new markets, significantly expanded the software business and elevated the collaboration with the ecosystem to a new level.

Jens Amail also played a key role in strengthening SNP’s corporate governance and strategic stability. The return to a dual governance structure with an Executive and Supervisory Board, as well as the resolution of long-standing legal matters, have provided a solid foundation for the company’s future development. On the product side, the introduction of the SNP Kyano platform in 2024 was an important evolutionary step in supporting the company’s strategy as an enabler of digital transformations and business agility.

Building on this strong momentum, SNP entered into a strategic partnership with global investment firm Carlyle at the end of 2024. The partnership is intended to support the next phase of the company’s growth by accelerating strategy execution and further expanding SNP’s global footprint.

With the early extension of Jens Amail’s contract, the Supervisory Board is ensuring continuity at the top of the company as SNP enters the next chapter of its growth story.

Hashtag: #SNP

The issuer is solely responsible for the content of this announcement.

About SNP

SNP (ticker: SHF.DE) is the global technology platform leader and trusted partner for companies seeking unparalleled data-enabled transformation capabilities and business agility. SNP’s Kyano platform integrates all necessary capabilities and partner offerings to provide a comprehensive software-based experience in data migration and management. Combined with the BLUEFIELD approach, Kyano sets a comprehensive industry standard for restructuring and modernizing SAP-centric IT landscapes faster and more securely while harnessing data-driven innovations.

The company works with more than 3,000 customers of all sizes and in all industries in over 80 countries, including numerous DAX 40 and Fortune 500 companies. The SNP Group has more than 1,600 employees worldwide at over 36 locations in 23 countries. The company is headquartered in Heidelberg, Germany, and generated revenues of EUR 254,8 million in the 2024 fiscal year.

More information is available at

Continue Reading

Trending