Results for “models”

MacRumors Mitchel Broussard Gadgets

Apple Watch Ultra 3 Gets Big $250 Discounts Ahead of Black Friday

Amazon and Best Buy today have a big collection of Apple Watch Ultra 3 models on sale, starting at $549.00, down from $799.00. This is a massive $250 discount on the previous-generation device, and it's a match for the all-time low price on these models. Note: MacRumors is an affiliate partner with some of these vendors. When you click a link and make a purchase, we may receive a small payment, which helps us keep the site running. You'll find both Black and Natural Apple…

Read at MacRumors

arXiv (research papers) Julian Kauk Tech policy

Large language models put existential risk from artificial intelligence at 10%

How likely is it that advanced artificial intelligence (AI) will cause human extinction or a similarly permanent disempowerment of humanity? Humans, including AI experts, frequently disagree on this question. Yet one important perspective is missing from this debate: that of AI systems themselves. We put the question to 241 large language models (LLMs) at every available reasoning setting. The models converged on a median risk of 10.0%, with the middle 90% of their estimates…

Read at arXiv (research papers)

9to5Mac Ryan Christoffel SI

tvOS 27.2.1 launching soon with Siri AI for older Apple TV 4K models

Today the new Apple TV 4K has arrived (order here) with Siri AI as a major tentpole feature. But older Apple TV 4K models are about to get Siri AI too via a tvOS 27.2.1 software update.

Read at 9to5Mac

South China Morning Post Chong Ming Lee,Minxiao Chang Geopolitics

Will Jensen Huang’s downfall tip the scales against open-source AI?

Jensen Huang’s abrupt resignation from Nvidia has cast uncertainty over the US chip giant’s multibillion-dollar investments in open-source artificial intelligence, while potentially depriving Chinese developers of one of their most influential advocates in Washington. His departure could prompt Nvidia to scale back spending on open-weight models and their supporting infrastructure, analysts said, although strong commercial incentives would likely keep the company committed…

Read at South China Morning Post

arXiv (research papers) Yundi Cai

ESI: An Embodied Self-Improvement Flywheel for LLM-VLA Agents

Independently pretrained language models and vision-language-action (VLA) policies provide complementary building blocks for embodied agents, but connecting them does not automatically produce coordinated behavior. The VLA may generate actions that diverge from the LLM's intended subtask and, upon regaining control after a correction, undo the promising state restored by the LLM. We introduce ESI (Embodied Self-Improvement), an automated flywheel that uses environmental…

Read at arXiv (research papers)

arXiv (research papers) Jun Miura

Evaluating Language Models as Spatial Priors for Object-Goal Navigation

Semantic exploration methods for object-goal navigation use large language models (LLMs) to score how strongly a target object is associated with a room type or with a nearby object. This paper considers obtaining from the LLM a spatial prior instead: a distribution over the position of the target in a room. We evaluate whether two open-weight LLMs, Qwen3.5-27B and Gemma-4-31B-it, can provide such a prior. Using real room layouts from CubiCasa5K and ScanNet, we measure the…

Read at arXiv (research papers)

Digital Trends Sudhanshu Kumar Mangalam

Samsung might be the unexpected reason Apple delayed the iPhone 18

Apple made a rather unusual change to its iPhone launch schedule this year. The company finally introduced its first foldable iPhone, the iPhone Duo, alongside the iPhone 18 Pro models. But the regular iPhone 18, which has traditionally launched alongside the Pro models, was nowhere to be seen. Earlier reports suggested Apple wanted to give

Read at Digital Trends

Kyiv Post Kyiv Post

Russia’s Starlink Rival Lags Behind Plans, Kyiv Says

Russia’s Rassvet satellites remain below their planned orbital altitude, Ukraine’s military intelligence said, while Moscow prepares more spacecraft and terminal models. Air Assault Forces commander Oleh Apostol said Russia could eventually overcome its technological difficulties, although development was taking longer than planned. The network already provides intermittent military communications over Ukraine, according to intelligence officials.

Read at Kyiv Post

Analytics India Magazine Kileuna Haralu SI

Adobe, Google Expand AI Partnership Across Creative and Productivity Tools

Adobe will integrate Google's AI models into its creative applications and bring Acrobat and other tools into Google's productivity and AI ecosystem.

Read at Analytics India Magazine

arXiv (research papers) Xianxu Wu

A Generalizable Approach to Safety-Aware Robotic Behavior Planning with an Explicit Constraint Space

Large language and vision-language models have substantially advanced robotic behavior planning from natural-language instructions and visual observations. However, safe manipulation in tabletop environments remains challenging because the safety constraints required for a task depend on the surrounding scene and may not be explicitly specified in the instruction. We propose a safety-aware behavior planning framework based on an explicit Constraint Space that stores reusable…

Read at arXiv (research papers)

Analytics India Magazine Siddharth Jindal SI

A Practical Blueprint for an AI-Ready Enterprise

An AI-ready enterprise needs room to experiment, safeguards for sensitive workloads and the freedom to adopt new models without starting again.

Read at Analytics India Magazine

MacRumors Tim Hardwick Gadgets

Apple Launches Two Products Today

Apple is releasing two new products today, Friday, November 13: the all-new Apple HomeView and the fourth-generation Apple TV 4K, both of which have been available to pre-order since mid-October. Apple HomeView: Apple's new smart home display, powered by Apple Intelligence with Siri AI, for entertainment and managing the home. Available in tabletop and wall-mounted models, it can recognize up to six users by face, each with a personalized Home Screen. The HomeView is priced…

Read at MacRumors

MarkTechPost Michal Sutter SI

Google Releases TPU Sync: Open-Source KV Cache Transfer for TPUs

Google released TPU Sync, an open-source C++ library that moves KV caches and weights across TPU HBM, host DRAM, storage and network. Google reports over 90% lower KV transfer latency on 235B to 480B MoE models.

Read at MarkTechPost

arXiv (research papers) Sol Choi

Reusing Locomotion Training Recipes Across Humanoids via Body-Aligned Virtual Joints

Reinforcement learning has enabled robust humanoid locomotion, yet training recipes (rewards, architectures, hyperparameters) remain hard to reuse across robots with different native joint coordinates, since policies output native joint angles while action scales, bounds, gains, and penalties are defined in those robot-specific coordinates. Existing transmission models convert native targets into actuator commands but do not standardize the coordinate interface exposed to…

Read at arXiv (research papers)

VietnamPlus (Vietnamese state media) Geopolitics

Vietnam plans to go green by expanding ASEAN eco-schools model

At the launch ceremony, Deputy Minister of Agriculture and Environment Le Trong Yen said schools play an important role in shaping young people's knowledge, character, and habits. Daily practices such as saving electricity and water, sorting waste and caring for trees could help students understand the value of natural resources and their responsibility towards the environment, while spreading positive habits to families and communities. Vietnam's eco-school models…

Read at VietnamPlus (Vietnamese state media)

Analytics India Magazine Ridhika Basnet SI

Google Launches Gemini 4 Hydrogen for Faster Coding and AI Workflows

Discover how Google's Gemini 4 Hydrogen enhances coding efficiency and AI workflows, outperforming previous models in key benchmarks.

Read at Analytics India Magazine

arXiv (research papers) Qi Tan Security

Condensing Law of Security Information

Scaling laws indicate that the generalization error of a large language model (LLM) decreases monotonically with model size once data is sufficient, and the corollary that larger models outperform smaller ones is now widely accepted. We argue that this corollary fails in security: decisions there hinge on anomalous inputs, which are long-tailed (e.g., zero-day attacks) and therefore poorly served by likelihood fitting. Worse still, the capability brought about by scale…

Read at arXiv (research papers)

arXiv (research papers) Alexander Spangher Tech policy

DOCKET: A Pluralistic Benchmark for Evaluating AI-Assisted Policy Feedback in U.S. Rulemaking

As AI models advance and impact deliberative democracies, there is a growing need to assess their support of pluralistic viewpoints. Most pluralism benchmarks are insufficient for this: they are not grounded in diversity (i.e., stakeholders across different frames, skills and needs), quality (i.e., rigorous and technical debates), and impact (i.e., invested deliberators and real-world outcomes). To address this, we first (i) curate DOCKET, the most comprehensive dataset of…

Read at arXiv (research papers)

arXiv (research papers) Wenzhuo Xu Security

LLMs are Vulnerable to Malicious Requests with Indistinguishable Intent

Jailbreak attacks can circumvent the safety mechanisms of Large Language Models (LLMs) and lead to the generation of harmful or policy-violating content. We first examine eight jailbreak methods against an LLM-based filter and observe that requests resembling normal user queries tend to have higher filter pass rates than those containing obvious anomalous patterns. This observation motivates the hypothesis that resemblance to normal requests may contribute to jailbreak…

Read at arXiv (research papers)

The Japan Times

China’s access to Nvidia chips in doubt as curbs gain traction in U.S. Congress

Nvidia's advanced chips are widely used to develop AI models, making the company key in the U.S. race against China to build superior AI systems.

Read at The Japan Times

arXiv (research papers) Yu Zhou

Beyond Prediction Gain: LASR as a Metric for Action-Grounded Latents in World Models

World models that support planning, control, and interpretable reasoning require latent representations that genuinely encode action. Yet the large-scale video data from which such models are learned overwhelmingly lack action annotations. Latent-action models infer action-like variables from observation sequences, enabling action-aware learning from unlabeled videos or trajectories. A fundamental concern, however, remains: a latent that improves future prediction may simply…

Read at arXiv (research papers)

arXiv (research papers) Shunyang Huang

ERA: Error-Resistant Autonomous Object Navigation with Predicted Depth

Depth and 3D foundation models can predict geometry from RGB images, allowing zero-shot object navigation without a depth sensor. However, when predicted depth is mapped as if it came from a sensor, view-dependent errors accumulate in the map, remove exploration goals, and end the search before the agent can observe and correct them. We propose ERA, a training-free object navigation system that replaces the depth sensor with depth predicted from RGB images and poses. ERA…

Read at arXiv (research papers)

Viet Nam News (state-run daily)

Community cooperatives sweep Forest Ecopreneur 2026 pitch round

Three cooperatives run by ethnic communities won top honours at Forest Ecopreneur 2026 with models that tie forest protection to higher local incomes.

Read at Viet Nam News (state-run daily)

Quantum Computing Report Mohamed Abdel-Kareem

Kipu Quantum Launches Rimay Quantum Feature Extraction in IBM Quantum Functions Catalog

Kipu Quantum has launched its Rimay quantum feature extraction service on the IBM Quantum Functions Catalog, allowing IBM Quantum users to access quantum-enhanced machine learning for feature engineering. This service, which operates as a preprocessing layer for classical machine learning models, has demonstrated improved recall and AUC scores in credit risk classification. It's designed to provide enriched data matrices for use with standard frameworks like scikit-learn.

Read at Quantum Computing Report

arXiv (research papers) Nel Escher Tech policy

Auditing Chatbots for Incursion into Professional Legal Practice

The provision of legal advice has long been in the exclusive jurisdiction of licensed attorneys-a boundary that unauthorized practice of law (UPL) doctrine maintains to protect the public from incompetent counsel. We argue that this boundary is being redrawn, not by courts or bar authorities, but by frontier AI companies. While early chat models were trained to avoid dispensing legal advice, modern models have been reconfigured to reduce refusals and prioritize helpfulness…

Read at arXiv (research papers)

Tech in Asia Glenn Kaonang Startups

The end of an era at Nvidia

Jensen Huang is out at Nvidia. Meanwhile, we zoom in on the race Jev sparked among AI decision models.

Read at Tech in Asia

MacRumors Juli Clover Gadgets

Siri AI Coming to Older Apple TV 4K Models in tvOS 27.2.1

An upcoming tvOS 27.2.1 update will bring Siri AI to the 2nd- and 3rd-generation Apple TV 4K models, according to Apple. A support document outlining ‌Siri‌ AI requirements across Apple products says tvOS 27.2.1 supports ‌Siri‌ AI on the ‌Apple TV‌ 4K (2nd generation and later), but the software hasn't yet been released. While older ‌Apple TV‌ models will need tvOS 27.2.1 for ‌Siri‌ AI, Apple appears to have a special…

Read at MacRumors

Ukrinform

Ukrainian drones to replace Mavics pass testing, best models to be ordered for military

The Brave1 defense innovation cluster, together with Ukraine’s Ministry of Defense, has tested reconnaissance drones of the Mavic type from nine Ukrainian manufacturers to assess whether the aircraft meet the military’s actual needs and the conditions of the modern battlefield.

Read at Ukrinform

arXiv (research papers) Mariana Sosa Guzmán

Grounding Object-to-Object Hazard Reasoning into 3D Risk Fields for Task-Conditioned Safety

Safety in robotics does not solely depend on geometric constraints, but also on semantics: the given task, environment properties, and interactions between objects in the scene. We present a framework based on vision-language models (VLMs) to infer task- and scene-dependent safety constraints for robot manipulation. We provide the VLM a structured safety tool based on an RGB-D observation and a natural-language task instruction that first decides whether the task should be…

Read at arXiv (research papers)

arXiv (research papers) Deepu Krishnareddy

Talk to Act: A Synthetic Multi-Turn Dialogue Benchmark for Human Intent Extraction and Action Planning in Robotics

Multi-turn conversational human-robot interaction (HRI) continues to be a challenging problem in robotics, as many existing systems rely on rigid, command-driven interfaces that can be unintuitive for users. Recent advances in Large Language Models (LLMs) provide a promising direction, given their ability to process natural language, generate structured outputs, and support tool-based reasoning, all of which are relevant for interpreting user intent in robotic settings…

Read at arXiv (research papers)

Older stories