Core Perspective - Solana aims to establish itself as the execution layer for "internet capital markets" in Asia, facilitating online asset issuance, trading, borrowing, lending, and settlement without traditional financial intermediaries [1] Group 1: Event Insights - The Solana Accelerate APAC event in Hong Kong featured a focus on institutional themes, emphasizing payments, tokenization, and the necessary infrastructure for integrating traditional finance [2] - Discussions included topics such as SOL staking ETFs, digital asset trusts, stablecoin infrastructure, tokenized securities, and regulated exchange-traded products [2][3] Group 2: Industry Collaboration - Asset managers like Mirae Asset and ChinaAMC participated alongside infrastructure firms such as CME Group and Fireblocks, indicating a strong interest from traditional financial institutions in the blockchain ecosystem [3] - Payment systems were a significant focus, with sessions dedicated to compliant stablecoin infrastructure and cross-border applications, highlighting a shift towards real-world adoption rather than speculative trading [3] Group 3: Infrastructure and AI - The event showcased the intersection of blockchain settlement layers and AI applications, reinforcing Solana's emphasis on speed and scalability [4] - The overall sentiment was one of resilience and commitment to building infrastructure despite market downturns, with a focus on practical solutions [5][6] Group 4: Practical Challenges - Key discussions revolved around the scalability of stablecoins, compliance for institutional onboarding, and the metrics that matter for selling on-chain solutions to asset managers and banks [6] - The event highlighted the need for user-friendly wallets and robust tokenization infrastructure that can withstand regulatory scrutiny [6] Group 5: Market Sentiment - The prevailing attitude was that while Solana is not immune to market cycles, the builders on the platform are focused on what truly matters, regardless of market conditions [7]

Core Insights - DeepSeek-OCR's Vision-Text Compression (VTC) technology achieves a compression rate of up to 10 times, significantly reducing the cost of processing long texts with large models [2][7] - The introduction of VTCBench, a benchmark test developed by research teams from institutions like the Chinese Academy of Sciences, aims to evaluate the cognitive limits of models in visual space through tasks such as information retrieval, associative reasoning, and long-term memory [2][10] VTC Technology Overview - VTC paradigm transforms long documents into high-density 2D images, which are then converted into a limited number of visual tokens by a visual encoder, differing from traditional models that read thousands of pure text tokens [6] - The technology can achieve a token compression rate between 2 to 10 times, significantly lowering computational and memory costs during long text processing [7] VTCBench Benchmark - VTCBench systematically evaluates models' cognitive limits in visual space through three main tasks: 1. VTC-Retrieval: Tests the model's ability to find specific facts in a vast visual context [10] 2. VTC-Reasoning: Challenges the model to find facts through associative reasoning with minimal text overlap [10] 3. VTC-Memory: Simulates long dialogues to assess the model's ability to resist decay of temporal and structural information [10] VTCBench-Wild - VTCBench-Wild has been introduced to assess the robustness of models in complex real-world scenarios, incorporating 99 different rendering configurations [11] Cognitive Bottlenecks - Current visual language models (VLMs) may excel at OCR recognition, but their understanding of high-density information from VTC-compressed texts remains questionable [9] - Testing results show a significant "U-shaped curve" in model performance, indicating that while models can capture information at the beginning and end of documents, their understanding of facts in the middle deteriorates as document length increases [14][15] Industry Insights - Despite the efficiency gains from VTC, existing VLMs still perform significantly worse than pure text LLMs in complex reasoning and memory tasks [17] - The performance of models like Gemini-3-Pro in VTCBench-Wild demonstrates that VTC is a highly feasible path for large-scale long text processing, with its visual understanding capabilities nearly matching pure text benchmarks [17][18]

Core Insights - The article discusses the limitations of existing Machine Learning Language Models (MLLMs) in specialized fields such as surgery, industry, extreme sports, and animal perspectives, highlighting the need for a new evaluation benchmark called EgoCross [1][3][9]. Group 1: EgoCross Benchmark - EgoCross is the first cross-domain egocentric video question-answering benchmark, covering four high-value professional fields and providing nearly a thousand high-quality QA pairs [3][9]. - The benchmark includes both closed-book (CloseQA) and open-book (OpenQA) evaluation formats, addressing a significant gap in the assessment of MLLMs [3][9]. Group 2: Model Evaluation and Findings - The research team tested eight mainstream MLLMs, revealing significant cross-domain shortcomings, with the best models achieving less than 55% accuracy in CloseQA and under 35% in OpenQA for cross-domain scenarios [4][12]. - The study found that models performed well in everyday activities but saw a drastic drop in accuracy when applied to specialized fields, with a notable decline from 73.58% in daily activities to 43.14% in cross-domain scenarios [12][18]. Group 3: Task Types and Challenges - The benchmark assesses four core tasks: identification, localization, prediction, and counting, with 15 sub-tasks designed to evaluate model capabilities comprehensively [11][12]. - Prediction tasks, such as forecasting the next action, showed a more significant decline in performance compared to basic identification tasks [18]. Group 4: Improvement Strategies - The research explored three improvement methods: prompt learning, supervised fine-tuning (SFT), and reinforcement learning (RL), with RL showing the most significant performance enhancement, averaging a 22% increase in CloseQA accuracy [15][14]. - SFT demonstrated nearly a 20% performance boost in the industrial domain, indicating the potential for targeted model training [15]. Group 5: Future Directions - The findings provide valuable insights into the current capabilities and limitations of large models, suggesting directions for developing more generalized multimodal systems [16][17].

Project Updates - Allora Network's $ALLO token has been listed on Binance spot exchange [1] - Kraken and other exchanges have also announced the listing of $ALLO [1] - Allora Network has announced a collaboration with Alibaba Cloud to support prediction models and DeAI workloads [1] Tokenomics - Validators and participants need to stake $ALLO to earn rewards and reputation [1] - Staking offers an Annual Percentage Yield (APY) of 12% - 50% [1] - Initial circulating supply of $ALLO is 2005% with 150% allocated for HODL airdrops [1] Binance HODLer Airdrop - Binance HODLer airdrop for Allora ($ALLO) is live, allowing users to earn ALLO airdrops by subscribing to BNB-denominated Simple Earn products [1] - Users who subscribe to BNB Simple Earn or On-Chain Products between October 23, 2025, 08:00 and October 26, 2025, 07:59 (UTC+8) will receive airdrop allocations [1]

Partnerships - Sui 与阿里云合作推出 AI 编码助手，支持 Sui Move 开发者 [1] Technology & Development - AI 编码助手支持英语、中文和韩语 [1]

Technology & Development - Move is described as a powerhouse [1] - Move is multilingual, secure, and fast for shipping with Alibaba Cloud's AI coding assistant in ChainIDE [1]

AI Technology Evolution - AI technology has evolved from solving artificial problems with shaky technology to addressing real-world problems with advanced technology [2][3] - Computing power advancements significantly change the way of thinking and problem-solving, similar to the evolution of transportation from bikes to cars to airplanes [5][6] - The classification of AI into AI, AGI, and ASI is not particularly useful; it's more about continuous evolution and increasing capabilities [7][8] Robotics and AI - Robotics is becoming integrated with AI, where AI provides the "engine" for robotics, similar to how EV cars use different engines than diesel cars [10][12] - AI outputs can be deployed in robots, but robotics remains a distinct field with its own fundamental technologies [11] China's AI Landscape - China offers vast opportunities for AI exploration and experimentation, with many projects likely to disappear in the future, but this exploration is valuable [13] - The Chinese market serves as a crucial testbed for maturing new technologies, playing a vital role in technology development [15][16] - China's AI innovation cycle is fast, driven by a collective effort and a "start-stop" mindset, where different organizations contribute and compete, leading to rapid iteration [19][20][21] - The long-term nature of AI development means that short-term advantages are unlikely to create insurmountable barriers for others to catch up [22][23] Challenges and Opportunities in AI Model Building - Creativity is the biggest challenge in AI model building today, as foundation models are already good enough, and the focus should be on developing innovative applications [28][29][30] - There's a need to fund creative individuals to develop applications for existing AI models, rather than solely focusing on replicating existing applications like ChatGPT [30] Cloud Computing and AI - Cloud computing is a lasting business with the potential to last for 50-100 years, similar to the fundamental nature of electricity [35] - AI is now a significant customer for cloud computing, highlighting the interconnectedness of these technologies [36] - The combination of computing, data, and models has fundamentally changed the way businesses operate, leading to the rise of AI [38][39]

Core Insights - ZTE Corporation, in collaboration with China Mobile and Alibaba Cloud, received the GSMA GLOMO "Open Gateway Challenge" award for their AaaS Open Gateway solution, highlighting significant commercial adoption in key industry scenarios [1][2] Group 1: Collaboration and Development - China Mobile is a pioneer in the GSMA Open Gateway initiative, establishing the world's largest Open Gateway platform that enhances its services and expands application scenarios [2][3] - The collaboration has led to the development of the QoD (Quality on Demand) API, which successfully passed 63 technical tests and achieved commercial deployment [2][3] Group 2: Technological Integration - The capability exposure solution integrates telecommunications and fintech strengths, ensuring superior network quality for mobile payments through China Mobile's CT middleware platform [3] - Alipay integrates with the AaaS Open Gateway platform, utilizing Alibaba Cloud's API to enhance wireless network performance in areas with weak coverage, thus improving transaction times and reducing failure rates [4] Group 3: Future Directions - China Mobile plans to continue collaborating with industry partners to develop advanced capability APIs and application scenarios, aiming to drive 5G commercialization and global Open Gateway adoption [5] - The Open Gateway initiative, led by GSMA, aims to globalize network capability exposure through standardized APIs, fostering innovation and interoperability in the digital ecosystem [6] Group 4: Industry Recognition - The GLOMO Awards are recognized as the industry's most prestigious accolades, celebrating innovation and excellence in the mobile industry with a judging panel of over 260 global analysts and experts [7]