All Updates
Page 420 of 1662
June 8, 2026
Geometric Insights into Language Model Activation Steering
This research explores the geometric mechanics of activation steering, revealing that concepts are primarily encoded in angular structure while hidden-state norms ensure stability. The authors propose decoupling these components to improve the precision and predictability of model interventions.
DiBS: Diffusion-Informed Branch Selection for Constraint Satisfaction
DiBS is a novel approach that integrates diffusion models into symbolic solvers to optimize branch selection in constraint satisfaction problems like Sudoku. By using diffusion models as a guide, it significantly reduces search costs and backtracks on complex instances.
CrowdMath: New Dataset for Collaborative Mathematical Reasoning
CrowdMath is a new dataset featuring 164 expert-annotated chains of collaborative mathematical research discussions. It aims to bridge the gap between solving well-specified problems and the complex, iterative process of open-problem solving.
CARVE-Q: Quantum-Accelerated Certified Autonomous Driving Repair
CARVE-Q introduces a quantum-AI hybrid architecture for certified, rule-compliant autonomous driving maneuvers. By utilizing quantum minimum finding, it efficiently solves complex interaction repair lattices while maintaining classical safety verification.
AFSAT: High-Performance GPU-Accelerated Pseudo-Boolean SAT Solver
AFSAT is a fully-engineered GPU-accelerated solver for pseudo-Boolean satisfiability that leverages the JAX compiler for massively parallel continuous local search. It improves upon previous proof-of-concept models by addressing floating-point limitations and memory latency, achieving near-linear scaling across multiple accelerators.
AEGIS: A Backup Reflex System for Physical AI
AEGIS is a selective escalation method that uses lightweight probes to detect high-risk steps in robot manipulation. It switches control to a stronger policy only when necessary, significantly improving recovery rates in long-horizon tasks.
Nvidia: Leading and Reshaping the PC Era
This article discusses Nvidia's strategic vision for the PC market, emphasizing the transformation of personal computing through AI integration. It highlights how Nvidia aims to redefine the role of PCs in the modern era.
April 2026 AI Trends in Product Design and Engineering
This booklet compiles key news from MONOist regarding the integration of AI in product development, design, and analysis. It provides engineers with a curated overview of how AI is transforming industrial workflows.
iPhone Fold dummy unit surfaces ahead of 2026 launch
Leaked photos of a dummy unit for the rumored iPhone Fold have surfaced, providing a glimpse into Apple's potential design direction. The device is currently expected to launch in the fall of 2026.
WeRide unveils WRD 3.0 end-to-end autonomous driving solution
WeRide showcased its new L2++ end-to-end autonomous driving solution, WRD 3.0, at the Qualcomm Summit. The technology highlights the company's progress in integrating advanced AI architectures for intelligent driving.
Jensen Huang Meets South Korean Tech Leaders Amid Market Dip
Nvidia CEO Jensen Huang visited Seoul to meet with leaders from SK Group, LG Group, and Naver Corp. This high-level engagement occurs as South Korean stocks face pressure following a pullback in AI-related investments.
Jensen Huang Calls Tech Selloff a Buying Opportunity
Nvidia CEO Jensen Huang dismissed concerns over the recent tech stock selloff, characterizing it as a buying opportunity. He emphasized that the global buildout of artificial intelligence infrastructure is still in its early stages.
NVIDIA and LG Group Partner to Build AI Factory
NVIDIA and LG Group are collaborating to construct an AI factory designed to accelerate development in robotics, autonomous driving, and data center technologies. This infrastructure will enable LG to train, simulate, and deploy AI applications across its diverse business units.
JPMorgan Hires Nomuraβs International AI Strategy Chief
JPMorgan Chase is strengthening its internal AI capabilities by hiring the international head of AI strategy from Nomura Holdings. This move reflects the bank's broader effort to integrate AI to boost productivity across its operations.
LLMs are Evolving into Large Programming Models
The article questions whether the linguistic capabilities of LLMs are truly improving with recent iterations, suggesting a shift in focus toward coding and logic tasks. It highlights the growing trend of models being optimized primarily for programming performance.
Casio pivots to AI robotics with Moflin pet success
Casio is diversifying its business beyond watches by focusing on the AI pet robot 'Moflin'. The company has established a president-led task force to cultivate this new growth pillar.
Ascend 910C Completes Training of 1.6 Trillion Parameter Model
A consortium including Huawei and Shenzhen Hetao College has successfully utilized the Ascend 910C to perform full-parameter post-training on a 1.6 trillion parameter model. This marks a significant milestone for domestic AI computing capabilities in China.
Big Tech AI Battle for China's Gaokao Season
Tencent, Alibaba, Baidu, and ByteDance are aggressively integrating their AI models into Gaokao preparation and support services. This competition highlights the strategic importance of educational AI applications in the Chinese market.
JD.com and Tencent partner on AI agent integration
JD.com and Tencent are collaborating to integrate AI agents across their platforms. The initiative uses Agent-to-Agent (A2A) technology to connect JD's e-commerce capabilities with Tencent's user ecosystem and major hardware manufacturers.
Zhu Ri Project Achieves Wireless Power Transmission Milestone
Academician Duan Baoyan's team at Xidian University has reached a breakthrough in long-distance wireless power transmission. This development is a critical step toward realizing space-based solar energy systems.