All Updates
Page 544 of 1666
May 21, 2026
NASA to reveal Moon Base project updates
NASA announced a press conference on May 26 to discuss the 'Moon Base' project, focusing on long-term human presence, industrial partnerships, and future mission schedules. Elon Musk expressed strong support for the initiative.
Zhejiang University Alumni Startup Funding Wave
A wave of startups founded by alumni of Zhejiang University are announcing new funding rounds. This highlights the growing influence of academic networks in the Chinese tech ecosystem.
Supply Chain Risks in Agricultural Products
The article discusses the risks of chemical contamination in agricultural products like bayberries, warning that quality control failures can destroy entire industry chains.
Why restaurants are pushing group-buying vouchers online
Restaurants are increasingly digitizing their sales through group-buying vouchers to capture online traffic. This shift highlights the broader trend of traditional service industries leveraging digital platforms to optimize customer acquisition.
Unitree vs. DeepRobotics: Embodied AI Strategies Compared
This article analyzes the diverging paths of Chinese robotics firms Unitree and DeepRobotics, focusing on their shift from quadruped robots to humanoid AI and their respective market positioning in B2B vs. C2C sectors.
Youdao Open-Sources Ziyue 4 Multimodal and TTS Engines
NetEase Youdao has upgraded its Ziyue LLM to version 4.0, introducing full-modal capabilities. The company has officially open-sourced its core multimodal and text-to-speech (TTS) models.
CHALI Tea's Collapse: The Failure of Traffic-Driven Models
The 'Chinese Lipton' CHALI faces bankruptcy after a failed pivot to bottled tea, highlighting the limitations of venture-backed, traffic-dependent consumer brands.
Anna's Archive hit with $19.5M copyright penalty
Anna's Archive, a shadow library, was ordered to pay $19.5 million in damages following a default judgment. A New York federal court issued a permanent injunction requiring global domain registrars and service providers to cease support.
Anthropic Co-founder Predicts Nobel-Level AI Breakthroughs Within Year
Anthropic co-founder Jack Clark predicts that AI will assist in a Nobel prize-winning discovery within 12 months. He also anticipates the rise of AI-run companies generating millions and the emergence of self-improving AI systems by 2028.
ResearchArena: Are AI Agents Ready for Scientific Discovery?
ResearchArena evaluates off-the-shelf AI agents on their ability to conduct full research loops, from ideation to paper writing. The study reveals that while agents can produce plausible-looking manuscripts, they struggle significantly with experimental rigor and factual accuracy.
New Framework Optimizes Exploration Under Volatility and Stochasticity
Researchers introduced CAUSE, a framework that distinguishes between environmental volatility and stochasticity to optimize decision-making. The study demonstrates that these two types of uncertainty drive exploration in opposite directions, providing a more precise model for adaptive AI agents.
New ECA Framework Prevents Multimodal Agent Hallucination Exploits
Researchers introduced Evidence-Carrying Multimodal Agents (ECA), a framework that forces agents to validate visual claims against external verifiers before executing privileged actions. This approach effectively mitigates 'hallucination-to-action' conversion, where models perform unauthorized tasks based on false perceptions.
MOCHA: Optimizing LLM Agent Skills via Chebyshev Annealing
MOCHA is a new optimization framework that treats LLM skill development as a multi-objective problem. By using Chebyshev scalarization and exponential annealing, it outperforms traditional prompt optimizers in balancing task performance with platform constraints.
LLMs Revolutionize Survey Research and Missing Data Imputation
Researchers developed a five-stage LLM framework for survey workflows, introducing the A-TLM model to outperform traditional imputation methods in disaster-related datasets. The study demonstrates that grounding LLMs in causal knowledge graphs significantly reduces bias and improves data quality.
Formalizing Trust Calibration for Agentic Tool Use
This research formalizes trust calibration in agentic systems as a preference-learning problem. It introduces a policy gateway that uses Gaussian-process uncertainty to decide when an agent should act autonomously or request human approval.
Formalizing Agentic Knowledge Graph Affordances for AI Agents
This paper introduces the Agentic Affordance Profile (AAP), a new semantic layer designed to help AI agents select and utilize Knowledge Graphs more effectively. It addresses current limitations in metadata standards by providing a formal framework for capability descriptions and epistemic soundness.
Causal Analysis Reveals Causes of MLLM Modality Hallucinations
Researchers identified an imbalance in attention heads that causes multimodal models to prioritize erroneous text over visual evidence. They introduced MACI, an intervention method that suppresses hallucination-driving heads during inference to improve accuracy.
Audit Reveals Lack of Reproducibility in LLM Trading Agents
This research paper analyzes 77 studies on LLM-based trading agents, revealing a critical lack of standardized evaluation protocols and reproducibility. The authors highlight that most current agentic trading research fails to account for transaction costs, execution timing, or robust backtesting standards.
2026 China AI Application Panorama Report Released
The report highlights a shift in AI applications from simple 'chatting' to 'executing tasks' (Agentic AI). It analyzes market trends, including the rise of AI agents, cost-effective model APIs, and the commercialization of B2B/C2C AI workflows.
Elon Musk's interconnected companies pose risks in SpaceX IPO
The SpaceX IPO filing reveals deep financial and operational entanglements between Elon Musk's various ventures, including xAI, Tesla, and X. Analysts view this complex network of cross-company interactions as a significant risk factor for potential investors.