Northwestern Polytechnical University in Xi'an, Shaanxi province, has developed a manta ray-inspired #submersible# capable of operating at depths of up to two kilometers beneath the water surface.

Video AI turns out to be surprisingly bad at pointing to "when and where," even when it can answer the question itself. Title: OmniCompass: A Holistic Benchmark for Omni-Modal Video Grounding and Understanding URL: 💡 Overview Researchers from Hong Kong Polytechnic University and OPPO Research Institute introduce OmniCompass, a new holistic benchmark for video understanding, evaluating 50 models on temporal/spatial grounding and question answering on the same footing. 🎯 The problem Existing benchmarks cover either vision-only or audio-visual evidence, test either QA or grounding but rarely both, and had no way to check whether a model's predictions actually agree when the target is specified by text alone versus text plus a reference image. 🧪 The method Across 820 videos (80.6 hours, 7 domains, 52 subcategories), each clip gets a matched Text setting and Text+Image setting sharing the same target and ground truth. Two tracks, OmniCompass-VL for vision-language models and OmniCompass-AV for omni-modal models, each evaluate three unified tasks: temporal grounding (TG), spatial grounding (SG), and QA. A new Cross-Specification Consistency (CS) score measures how well predictions agree between the two settings. 📊 Results QA reaches up to 69.1% (VL) and 78.2% (AV), while temporal grounding tops out at just 27.1% (VL) and 40.6% (AV), exposing a huge gap between answering and pointing. The best open-source model also trails the best proprietary model, Gemini-3.8-Flash, by 32.8 points on CS average in the AV track (24.6% vs. 57.4%). #VideoAI# #Benchmarks#

Video AI turns out to be surprisingly bad at pointing to "when and where," even when it can answer the question itself. Title: OmniCompass: A Holistic Benchmark for Omni-Modal Video Grounding and Understanding URL: 💡 Overview Researchers from Hong Kong Polytechnic University and OPPO Research Institute introduce OmniCompass, a new holistic benchmark for video understanding, evaluating 50 models on temporal/spatial grounding and question answering on the same footing. 🎯 The problem Existing benchmarks cover either vision-only or audio-visual evidence, test either QA or grounding but rarely both, and had no way to check whether a model's predictions actually agree when the target is specified by text alone versus text plus a reference image. 🧪 The method Across 820 videos (80.6 hours, 7 domains, 52 subcategories), each clip gets a matched Text setting and Text+Image setting sharing the same target and ground truth. Two tracks, OmniCompass-VL for vision-language models and OmniCompass-AV for omni-modal models, each evaluate three unified tasks: temporal grounding (TG), spatial grounding (SG), and QA. A new Cross-Specification Consistency (CS) score measures how well predictions agree between the two settings. 📊 Results QA reaches up to 69.1% (VL) and 78.2% (AV), while temporal grounding tops out at just 27.1% (VL) and 40.6% (AV), exposing a huge gap between answering and pointing. The best open-source model also trails the best proprietary model, Gemini-3.8-Flash, by 32.8 points on CS average in the AV track (24.6% vs. 57.4%). #VideoAI# #Benchmarks#

Zlatan’s homecoming celebration is a reminder that dreams are valid and success can inspire generations. From the classrooms to the Moshood Abiola Polytechnic to the Spotlight 🔥

She's the Mexican scientist who just rewrote the rules on one of the world's most common viruses. After 20 years of relentless research at Mexico's Instituto Politécnico Nacional, Dr. Gallegos pulled off something the medical world hadn't seen before. She wiped out HPV. Completely. Her weapon? A non-invasive technique called photodynamic therapy — no scalpels, no chemo, no chemotherapy aftermath. Here's how it works. A drug called delta-aminolevulinic acid is applied to the cervix. It transforms into a fluorescent compound that only sticks to damaged cells. Healthy cells flush it out. Then a beam of light is aimed at the area — and the infected cells get destroyed on the spot. Healthy tissue? Untouched. The numbers are wild. In 29 women in Mexico City who had HPV without lesions, the virus was eliminated in 100% of cases. In the broader study across Oaxaca and Veracruz, 420 more women were treated — with 64.3% success in those carrying both HPV and premalignant lesions, and 57.2% in those with lesions alone. This matters more than people realise. HPV is the most common sexually transmitted infection on the planet. Around 80% of women will catch some form of it in their lifetime. It's behind nearly every case of cervical cancer worldwide. And until now, there was no cure. Only vaccines and screenings. Dr. Gallegos changed that conversation. Quietly. From a lab in Mexico City. Without the global spotlight she deserves. One woman. Two decades. A breakthrough that could save millions. Source: Instituto Politécnico Nacional (IPN), Mexico City — research led by Dr. Eva Ramón Gallegos

It looks like one of the Polish officers applied a tourniquet to his right arm. It may have saved his life. All British police officers should be issued and should be required to carry a Combat Application Tourniquet (CAT) and haemostatic gauze on them, a trauma kit in the car, and be properly trained in their use. Shockingly, that is not presently the case. @PFEW_HQ @PoliceChiefs @ukhomeoffice @ShabanaMahmood

Post-05 generation steps up: they started AI companies while still in school In March this year, several young people won the gold award in the public welfare track of an #AI# competition. Liu Sizhe, the youngest team member, is about to turn 18. He is founding a company to put the Alzheimer's AI early-screening project "Forget-Me-Not" on track, using AI to call and chat with elderly people to help with early detection and intervention. Mei Youshan from Zhejiang Gongshang University runs an overseas power tool supply chain platform and has taken over 1 million yuan in overseas orders. Tang Yanwei from the Communication University of China is working on an AI education product "UniGrow," passing on experience through "top-student AI avatars." Hou Zenghao from Shenzhen Polytechnic University is working on "aerial intelligent agent" drone inspections.

South Korea’s Vinatech has secured KRW242.5 billion in data-center supercapacitor orders in just three days—nearly three times its entire 2025 revenue, The Elec reports. The latest deal, disclosed on Oct. 21, is a KRW119.5 billion contract with Sanmina covering supplies to India through December 2027. It follows a KRW123 billion order from Taiwan’s AcBel Polytech announced on Oct. 19. Supercapacitors can supply or absorb energy in short bursts, helping smooth the sharp swings in power demand from AI servers. The Elec puts Vinatech’s disclosed data-center orders since late June at about KRW348.5 billion, more than four times last year’s revenue. Customers include AcBel, Sanmina and Bloom Energy.