<img width="980" height="490" src="https://insideainews.com/wp-content/uploads/2025/11/google-cloud-ironwood-pod-2-1-112025.png" class="webfeedsFeaturedVisual wp-post-image" alt="" style="float: right; margin-left: 5px;" link_thumbnail="" decoding="async" srcset="https://insideainews.com/wp-content/uploads/2025/11/google-cloud-ironwood-pod-2-1-112025.png 980w, https://insideainews.com/wp-content/uploads/2025/11/google-cloud-ironwood-pod-2-1-112025-150x75.png 150w, https://insideainews.com/wp-content/uploads/2025/11/google-cloud-ironwood-pod-2-1-112025-300x150.png 300w, https://insideainews.com/wp-content/uploads/2025/11/google-cloud-ironwood-pod-2-1-112025-768x384.png 768w" sizes="(max-width: 980px) 100vw, 980px" />Google today announced GA on the Google Cloud Platform of three products built on custom silicon built for inference and agentic workloads: &#8211; Ironwood, Google’s seventh generation Tensor Processing Unit, will be generally available in the coming weeks. The company said it is built for large-scale model training and complex reinforcement learning, as well as [&#8230;]

أعلنت Google Cloud عن توفر وحدات معالجة Tensor Ironwood بشكل عام، المصممة لتدريب النماذج على نطاق واسع والتعلم المعزز المعقد. يمثل هذا تقدمًا كبيرًا في بنية الذكاء الاصطناعي، مما يتيح للمطورين الاستفادة من قدرات حسابية محسّنة لتطبيقاتهم.

Google Cloud ha anunciado la disponibilidad general de sus Unidades de Procesamiento Tensor Ironwood, diseñadas para el entrenamiento de modelos a gran escala y el aprendizaje por refuerzo complejo. Esto representa un avance significativo en la infraestructura de IA, permitiendo a los desarrolladores aprovechar capacidades computacionales mejoradas para sus aplicaciones.

Google Cloud a annoncé la disponibilité générale de ses unités de traitement Tensor Ironwood, conçues pour l'entraînement de modèles à grande échelle et l'apprentissage par renforcement complexe. Cela représente une avancée significative dans l'infrastructure de l'IA, permettant aux développeurs de tirer parti de capacités de calcul améliorées pour leurs applications.

Google Cloud has announced the general availability of its Ironwood Tensor Processing Units (TPUs), designed for large-scale model training and complex reinforcement learning. This marks a significant advancement in AI infrastructure, allowing developers to leverage enhanced computational capabilities for their applications.

Google Cloud Announces GA of Ironwood TPUs

arXiv:1906.07172v5 Announce Type: replace-cross 
Abstract: Equivariant neural networks are a class of neural networks designed to preserve symmetries inherent in the data. In this paper, we introduce a general method for modifying a neural network to enforce equivariance, a process we refer to as equivarification. We further show that group convolutional neural networks (G-CNNs) arise as a special case of our framework.

Equivariant neural networks and equivarification

arXiv:2311.13745v4 Announce Type: replace-cross 
Abstract: Diffusion models have become the most popular approach to deep generative modeling of images, largely due to their empirical performance and reliability. From a theoretical standpoint, a number of recent works have studied the iteration complexity of sampling, assuming access to an accurate diffusion model. In this work, we focus on understanding the sample complexity of training such a model; how many samples are needed to learn an accurate diffusion model using a sufficiently expressive neural network? Prior work showed bounds polynomial in the dimension, desired Total Variation error, and Wasserstein error. We show an exponential improvement in the dependence on Wasserstein error and depth, along with improved dependencies on other relevant parameters.

Improved Sample Complexity Bounds for Diffusion Model Training

arXiv:2410.22995v2 Announce Type: replace-cross 
Abstract: A hallmark of advanced artificial intelligence is the capacity to progress from passive visual perception to the strategic modification of visual information to facilitate complex reasoning. This advanced capability, however, remains critically underdeveloped in current Large Multi-modal Models (LMMs). The deficiency is often masked by evaluation metrics that prioritize final-answer accuracy, creating an illusion of competence where genuine reasoning is absent. Using the domain of geometric problem-solving as a precise instrument, we probe this issue through tasks that require constructing visual aids. To this end, we introduce \textbf{VisAidMath}, a challenging benchmark, and our novel Three-Layered Funnel Evaluation Framework. This framework moves beyond simple accuracy (ACCU) to scrutinize the generation of valid visual aids (PVA) and the soundness of subsequent reasoning steps (SPRS). Our extensive experiments on state-of-the-art models, including Doubao-Seed-1.6 and o4, reveal a profound ``Reasoning Illusion''. We observe that high surface-level accuracy conceals a catastrophic failure in the models' ability to produce valid visual aids or to reason from them. Our findings expose a fundamental schism between visual perception and logical deduction in modern LMMs. We host an evaluation platform at CodaBench for testing publicly. Homepage: https://nlp2ct.github.io/VisAidMathHomepage/ Evaluation: https://www.codabench.org/competitions/7634/

VisAidMath: Benchmarking Visual-Aided Mathematical Reasoning

arXiv:2508.01249v3 Announce Type: replace-cross 
Abstract: Large Language Model (LLM) agents offer a powerful new paradigm for solving various problems by combining natural language reasoning with the execution of external tools. However, their dynamic and non-transparent behavior introduces critical security risks, particularly in the presence of prompt injection attacks. In this work, we propose a novel insight that treats the agent runtime traces as structured programs with analyzable semantics. Thus, we present AgentArmor, a program analysis framework that converts agent traces into graph intermediate representation-based structured program dependency representations (e.g., CFG, DFG, and PDG) and enforces security policies via a type system. AgentArmor consists of three key components: (1) a graph constructor that reconstructs the agent's runtime traces as graph-based intermediate representations with control and data flow described within; (2) a property registry that attaches security-relevant metadata of interacted tools \& data, and (3) a type system that performs static inference and checking over the intermediate representation. By representing agent behavior as structured programs, AgentArmor enables program analysis for sensitive data flow, trust boundaries, and policy violations. We evaluate AgentArmor on the AgentDojo benchmark, the results show that AgentArmor can reduce the ASR to 3\%, with the utility drop only 1\%.

AgentArmor: Enforcing Program Analysis on Agent Runtime Trace to Defend Against Prompt Injection

arXiv:2511.14691v1 Announce Type: cross 
Abstract: Attention is the brain's ability to selectively focus on a few specific aspects while ignoring irrelevant ones. This biological principle inspired the attention mechanism in modern Transformers. Transformers now underpin large language models (LLMs) such as GPT, but at the cost of massive training and inference energy, leading to a large carbon footprint. While brain attention emerges from neural circuits, Transformer attention relies on dot-product similarity to weight elements in the input sequence. Neuromorphic computing, especially spiking neural networks (SNNs), offers a brain-inspired path to energy-efficient intelligence. Despite recent work on attention-based spiking Transformers, the core attention layer remains non-neuromorphic. Current spiking attention (i) relies on dot-product or element-wise similarity suited to floating-point operations, not event-driven spikes; (ii) keeps attention matrices that suffer from the von Neumann bottleneck, limiting in-memory computing; and (iii) still diverges from brain-like computation. To address these issues, we propose the Spiking STDP Transformer (S$^{2}$TDPT), a neuromorphic Transformer that implements self-attention through spike-timing-dependent plasticity (STDP), embedding query--key correlations in synaptic weights. STDP, a core mechanism of memory and learning in the brain and widely studied in neuromorphic devices, naturally enables in-memory computing and supports non-von Neumann hardware. On CIFAR-10 and CIFAR-100, our model achieves 94.35\% and 78.08\% accuracy with only four timesteps and 0.49 mJ on CIFAR-100, an 88.47\% energy reduction compared to a standard ANN Transformer. Grad-CAM shows that the model attends to semantically relevant regions, enhancing interpretability. Overall, S$^{2}$TDPT illustrates how biologically inspired attention can yield energy-efficient, hardware-friendly, and explainable neuromorphic models.

Attention via Synaptic Plasticity is All You Need: A Biologically Inspired Spiking Neuromorphic Transformer

arXiv:2511.08955v2 Announce Type: replace-cross 
Abstract: Simulating microstructure evolution (MicroEvo) is vital for materials design but demands high numerical accuracy, efficiency, and physical fidelity. Although recent studies on deep learning (DL) offer a promising alternative to traditional solvers, the field lacks standardized benchmarks. Existing studies are flawed due to a lack of comparing specialized MicroEvo DL models with state-of-the-art spatio-temporal architectures, an overemphasis on numerical accuracy over physical fidelity, and a failure to analyze error propagation over time. To address these gaps, we introduce MicroEvoEval, the first comprehensive benchmark for image-based microstructure evolution prediction. We evaluate 14 models, encompassing both domain-specific and general-purpose architectures, across four representative MicroEvo tasks with datasets specifically structured for both short- and long-term assessment. Our multi-faceted evaluation framework goes beyond numerical accuracy and computational cost, incorporating a curated set of structure-preserving metrics to assess physical fidelity. Our extensive evaluations yield several key insights. Notably, we find that modern architectures (e.g., VMamba), not only achieve superior long-term stability and physical fidelity but also operate with an order-of-magnitude greater computational efficiency. The results highlight the necessity of holistic evaluation and identify these modern architectures as a highly promising direction for developing efficient and reliable surrogate models in data-driven materials science.

MicroEvoEval: A Systematic Evaluation Framework for Image-Based Microstructure Evolution Prediction

<img width="352" height="176" src="https://insideainews.com/wp-content/uploads/2025/05/cloudera-logo-2-1-0525.png" class="webfeedsFeaturedVisual wp-post-image" alt="" style="float: right; margin-left: 5px;" link_thumbnail="" decoding="async" fetchpriority="high" srcset="https://insideainews.com/wp-content/uploads/2025/05/cloudera-logo-2-1-0525.png 352w, https://insideainews.com/wp-content/uploads/2025/05/cloudera-logo-2-1-0525-150x75.png 150w, https://insideainews.com/wp-content/uploads/2025/05/cloudera-logo-2-1-0525-300x150.png 300w" sizes="(max-width: 352px) 100vw, 352px" />By Abhas Ricky, Chief Strategy Officer at Cloudera Artificial intelligence drives the next wave of enterprise transformation, yet many organizations remain stuck. Concerns about keeping sensitive data and intellectual property secure are holding enterprises back from AI adoption. According to a recent Accenture study, 77% of organizations lack the foundational data and AI security practices [&#8230;]

يتحدث أبهاس ريكي، المدير التنفيذي للاستراتيجية في شركة كلوديرا، عن الإمكانات التحويلية للذكاء الاصطناعي (AI) في المؤسسات. ومع ذلك، فإن العديد من المنظمات مترددة في اعتماد الذكاء الاصطناعي بسبب المخاوف المتعلقة بأمان البيانات الحساسة والملكية الفكرية. تشير دراسة حديثة أجرتها أكسنتشر إلى أن 77% من المنظمات تفتقر إلى الممارسات الأساسية لأمان البيانات والذكاء الاصطناعي، مما يعيق قدرتها على الاستفادة من الذكاء الاصطناعي بشكل فعال.

Abhas Ricky, director de estrategia de Cloudera, discute el potencial transformador de la inteligencia artificial (IA) para las empresas. Sin embargo, muchas organizaciones son reacias a adoptar la IA debido a preocupaciones sobre la seguridad de los datos sensibles y la propiedad intelectual. Un estudio reciente de Accenture indica que el 77% de las organizaciones carecen de las prácticas de seguridad de datos y de IA necesarias, lo que obstaculiza su capacidad para aprovechar la IA de manera efectiva.

Abhas Ricky, directeur de la stratégie chez Cloudera, aborde le potentiel transformateur de l'intelligence artificielle (IA) pour les entreprises. Cependant, de nombreuses organisations hésitent à adopter l'IA en raison de préoccupations concernant la sécurité des données sensibles et de la propriété intellectuelle. Une étude récente d'Accenture indique que 77 % des organisations manquent des pratiques de sécurité des données et de l'IA nécessaires, ce qui entrave leur capacité à tirer parti de l'IA de manière efficace.

Abhas Ricky, Chief Strategy Officer at Cloudera, discusses the transformative potential of artificial intelligence (AI) for enterprises. However, many organizations are hesitant to adopt AI due to concerns about the security of sensitive data and intellectual property. A recent Accenture study indicates that 77% of organizations lack the necessary foundational data and AI security practices, which hinders their ability to leverage AI effectively.

Private AI: Moving Models to Data within Secure Boundaries

<img width="457" height="225" src="https://insideainews.com/wp-content/uploads/2025/11/qumulo-logo-2-1-0324.png" class="webfeedsFeaturedVisual wp-post-image" alt="" style="float: right; margin-left: 5px;" link_thumbnail="" decoding="async" srcset="https://insideainews.com/wp-content/uploads/2025/11/qumulo-logo-2-1-0324.png 457w, https://insideainews.com/wp-content/uploads/2025/11/qumulo-logo-2-1-0324-150x74.png 150w, https://insideainews.com/wp-content/uploads/2025/11/qumulo-logo-2-1-0324-300x148.png 300w" sizes="(max-width: 457px) 100vw, 457px" />SEATTLE &#8211; Nov. 12, 2025 — Unstructured data management company Qumulo today unveiled three products — Qumulo Helios Agent, Qumulo Cloud AI Accelerator and Qumulo AI Networking — designed to support how enterprises integrate their data into AI factories. “Enterprises in today’s competitive environments need more than just storage—they need systems that think, adapt, and [&#8230;]

في 12 نوفمبر 2025، أعلنت شركة كويمولو، المتخصصة في إدارة البيانات غير المنظمة، عن إطلاق ثلاثة منتجات جديدة: وكيل كويمولو هيليوس، مسرع كويمولو السحابي للذكاء الاصطناعي، وشبكة كويمولو للذكاء الاصطناعي. تم تصميم هذه المنتجات لتحسين كيفية إدارة الشركات لبياناتها ودمجها في مصانع الذكاء الاصطناعي، مما يلبي الحاجة المتزايدة إلى أنظمة إدارة بيانات متقدمة يمكنها التكيف والتفكير في البيئات التنافسية.

El 12 de noviembre de 2025, Qumulo, una empresa especializada en la gestión de datos no estructurados, anunció el lanzamiento de tres nuevos productos: Qumulo Helios Agent, Qumulo Cloud AI Accelerator y Qumulo AI Networking. Estos productos están diseñados para mejorar cómo las empresas gestionan e integran sus datos en fábricas de IA, abordando la creciente necesidad de sistemas avanzados de gestión de datos que puedan adaptarse y pensar en entornos competitivos.

Le 12 novembre 2025, Qumulo, une entreprise spécialisée dans la gestion des données non structurées, a annoncé le lancement de trois nouveaux produits : Qumulo Helios Agent, Qumulo Cloud AI Accelerator et Qumulo AI Networking. Ces produits visent à améliorer la manière dont les entreprises gèrent et intègrent leurs données dans les usines d'IA, répondant ainsi à la nécessité croissante de systèmes de gestion des données avancés capables de s'adapter et de réfléchir dans des environnements compétitifs.

On November 12, 2025, Qumulo, a company specializing in unstructured data management, announced the launch of three new products: Qumulo Helios Agent, Qumulo Cloud AI Accelerator, and Qumulo AI Networking. These products are designed to enhance how enterprises manage and integrate their data within AI factories, addressing the growing need for advanced data management systems that can adapt and think in competitive environments.

Qumulo Announces Global Data Supply Chain for AI Factories

<img width="422" height="211" src="https://insideainews.com/wp-content/uploads/2025/04/mlcommons-logo-2-1-1124.png" class="webfeedsFeaturedVisual wp-post-image" alt="" style="float: right; margin-left: 5px;" link_thumbnail="" decoding="async" srcset="https://insideainews.com/wp-content/uploads/2025/04/mlcommons-logo-2-1-1124.png 422w, https://insideainews.com/wp-content/uploads/2025/04/mlcommons-logo-2-1-1124-150x75.png 150w, https://insideainews.com/wp-content/uploads/2025/04/mlcommons-logo-2-1-1124-300x150.png 300w" sizes="(max-width: 422px) 100vw, 422px" />Today, MLCommons announced new results for the MLPerf Training v5.1 benchmark suite, highlighting the rapid evolution and increasing richness of the AI ecosystem as well as significant performance improvements from new generations of systems. Go here to view the full results for MLPerf Training v5.1 and find additional information about the benchmarks. The MLPerf Training benchmark suite comprises [&#8230;]

أعلنت MLCommons عن نتائج مجموعة معايير MLPerf Training v5.1، مما يبرز التطور السريع والغنى المتزايد لنظام الذكاء الاصطناعي. تسلط النتائج الجديدة الضوء على تحسينات كبيرة في الأداء من الأجيال الجديدة من الأنظمة. يبرز هذا الإصدار التقدم في تكنولوجيا الذكاء الاصطناعي وتأثيرها المتزايد على مختلف الصناعات.

MLCommons ha anunciado los resultados de la suite de benchmarks MLPerf Training v5.1, destacando la rápida evolución y el creciente enriquecimiento del ecosistema de IA. Los nuevos resultados subrayan mejoras significativas en el rendimiento de las últimas generaciones de sistemas. Este lanzamiento subraya los avances en la tecnología de IA y su creciente impacto en diversas industrias.

MLCommons a annoncé les résultats de la suite de benchmarks MLPerf Training v5.1, mettant en avant l'évolution rapide et la richesse croissante de l'écosystème de l'IA. Les nouveaux résultats soulignent des améliorations significatives des performances des dernières générations de systèmes. Cette publication souligne les avancées de la technologie IA et son impact croissant sur divers secteurs.

MLCommons has announced the results of the MLPerf Training v5.1 benchmark suite, showcasing the rapid evolution and increasing richness of the AI ecosystem. The new results highlight significant performance improvements from the latest generations of systems. This release underscores the advancements in AI technology and its growing impact on various industries.

Google Cloud Announces GA of Ironwood TPUs

Was this article worth reading? Share it