- 资讯公开站Digitimes
FAW, GAC formalize partnership as Toyota's China joint ventures face possible overhaul
China's state-owned FAW and GAC's controlling shareholder, Guangzhou Automobile Industry Group, have signed a strategic cooperation framework that could reshape Toyota's long-standing two-joint-venture model in the count…
- 资讯公开站Digitimes
Hygon, Sugon deepen integration under one chairman after US$17.3B merger collapse
Chinese AI chipmaker Hygon Information Technology has appointed Jun Li as its new chairman following a recent board reshuffle. Li only took over as chairman of Chinese server maker Dawning Information Industry (Sugon) in…
- 资讯公开站Digitimes
Samsung launches Tab S12, Buds On, SmartTag3
Samsung Electronics on October 1 unveiled its next-generation premium Galaxy Tab S12 series, its first clip-on wireless earbuds, the Galaxy Buds On, and the Galaxy SmartTag3, which expands iOS support. The three product …
- 资讯公开站Digitimes
Chinese automakers' shift to the US may reshape supply chains and supplier margins
Hiroca said two major Chinese automakers are likely to build factories in the US, a move that could redraw supply chains across North America, alter supplier demand patterns, and deepen pressure on parts makers already f…
- 资讯公开站Digitimes
SK Hynix says Solidigm IPO remains undecided as AI storage demand drives investment
SK Hynix has reiterated that no decision has been made on a potential US initial public offering (IPO) or other external financing for its Solidigm subsidiary, as the company weighs investment needs driven by growing dem…
- 资讯公开站Digitimes
Wi-Fi 8 races ahead of the standard, putting networking chips on a faster 2027 growth track
Wi-Fi 8 adoption is accelerating well ahead of the standard's expected finalization in the first half of 2028, with device brands and chipmakers already moving into early commercial deployment. The first Wi-Fi 8 routers …
- 资讯公开站Digitimes
Toshiba reportedly to double HDD production as AI drives storage demand
Toshiba plans to double production capacity for hard disk drives (HDDs) used in AI data centers by fiscal 2027, investing about JPY60 billion (US$379.4 million) to expand its Philippine operations, according to Nikkei As…
- 资讯公开站Digitimes
Meta hires MongoDB CEO to lead AI enterprise platform
Meta has launched a new business unit, "Meta Enterprise Platform," to package its artificial intelligence (AI) models, agent tools, and infrastructure for enterprise customers, and has tapped MongoDB chief executive CJ D…
- 资讯公开站Digitimes
Climax Technology sets up AI center, targets 3Q27 sampling
Wireless communications company Climax Technology is accelerating its AI push by launching an "AI and Advanced Technology Center," with its first project a next-generation AI sensing platform slated for sampling in the t…
- 资讯公开站Digitimes
Southern Taiwan Science Park approves three new companies, pushing 2026 investment over NT$100 billion
The Southern Taiwan Science Park (STSP) has approved the entry of three new companies—Fangsco, Nanya Technology, and South Korea's Hansol Chemical—pushing the park's cumulative investment to more than NT$100 …
- 资讯公开站Digitimes
Swancor expands recycled materials, eyes recyclable carbon fiber for drone components
Swancor Holding subsidiary Swancor Regener is jointly exhibiting with Aeraxa Aerospace Technology at the Taichung AI and Smart Manufacturing Show in Taichung, Taiwan, showcasing products and technologies including recycl…
- 资讯公开站Digitimes
Onsemi revises Synaptics takeover deal after rival bid
Onsemi has amended its agreement to buy Synaptics in an all-cash deal, a move that could reshape semiconductor competition and affect global technology supply chains. The revised cash transaction offers Synaptics shareho…
- 论文公开站arXiv
Generative modeling of intrinsically disordered protein regions by reinforcing sparse autoencoder features
Intrinsically disordered protein regions (IDRs) play central roles in cellular processes such as transcriptional regulation, signal transduction, and subcellular localization, yet their functional design remains challeng…
- 论文公开站arXiv
Trust the Direction, Search the Step: Zero-and-First-Order Methods for LLM Fine-Tuning
Step-size selection remains a central challenge in large-scale neural network optimization; conservative steps slow convergence, while aggressive steps can destabilize it. We combine \textbf{Z}ero-and-\textbf{F}irst-\tex…
- 论文公开站arXiv
The Missing Primitive: Diagnosing and Repairing Mathematical Reasoning in Large Language Models
While Large Language Models (LLMs) have demonstrated striking capabilities on frontier mathematical problems, it remains unclear whether they possess the structural mathematical understanding underlying their solutions. …
- 论文公开站arXiv
Hierarchical Continuous Diffusion Language Models
Discrete diffusion language models offer a compelling alternative to autoregressive generation for tasks demanding bidirectional reasoning and global constraint satisfaction. Yet they share a structural bottleneck: when …
- 论文公开站arXiv
Cost-augmented Schrödinger bridges on graphs are exactly solvable: a Feynman-Kac tilt replaces learned control
The generalized Schrödinger bridge on a graph moves mass between two distributions while charging a cost for the states visited. It has been approached by learning the rates of a controlled continuous-time Markov chain, …
- 论文公开站arXiv
InterEvolve: Test-Time Evolution of Reward Programs for Humanoid Loco-Manipulation
We study test-time evolution for humanoid loco-manipulation: solving tasks that a controller was never trained for by repurposing its existing skills, improving from its own attempts, and retaining what it learns, withou…
- 论文公开站arXiv
FERPO: Forward Entropy-Regularized Policy Optimization
Several state-of-the-art methods for online reinforcement learning in continuous control improve policies using action gradients of a learned critic. However, critics are typically trained to predict returns, and accurat…
- 论文公开站arXiv
TACO: Ternary Absolute-max Column-wise One-sparse Optimizer for LLM Fine-Tuning
Full-parameter fine-tuning of large language models (LLMs) incurs substantial optimizer state memory overhead, limiting the model sizes that fit on modern GPUs. Existing approaches either compress optimizer state, abando…
- 论文公开站arXiv
VISTA: A Visual Harness for Reasoning in an Interactive World
We show that multimodal models possess strong reasoning abilities and that an appropriate harness can unlock their potential to solve tasks across diverse interactive environments. We introduce VISTA, a visual harness th…
- 论文公开站arXiv
SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation
High-resolution 3D generation increasingly relies on voxel latents and multi-stage pipelines that first predict active structure and then synthesize local geometry. While effective, this design fragments continuous surfa…
- 论文公开站arXiv
ScholarCatalyst: A Benchmark for Retrieving Papers That Inspire New Research
What makes great scientists great? Even as AI systems start to make progress on open problems, scientists remain far ahead of them at sensing which prior idea, buried in an ever-growing archive of research, a new problem…
- 论文公开站arXiv
Embedding Prediction Helps Image Generation
In diffusion transformers, a class label or a text prompt is embedded once, and the same condition is reused at every denoising step. We ask whether predicted embeddings can serve as this condition instead. Next-Embeddin…
- 论文公开站arXiv
Reconstruct, Practice, Go Real: Guided Self-Improvement for Embodied Agents
Building reliable robot capabilities across diverse tasks requires substantial human effort to develop and maintain skills, design rewards, and integrate perception with control. We present Reconstruct, Practice, Go Real…
