AI News Archive: August 12, 2026 — Part 15
Sourced from 500+ daily AI sources, scored by relevance.
- IBM Signs $240M Deal for Nvidia-Powered AI Cluster
The partnership with Together AI comes amid booming demand for open models.
- IBM finds a neocloud cash injection with $240m Together AI deal
The multi-year deal will see the neocloud rent a fresh batch of NVIDIA servers.
- IBM and Together AI sign $240m multi-year cloud agreement
IBM has struck a multi-year agreement worth $240m with Together AI, to build a large cluster of Nvidia HGX B300 systems on IBM Cloud.
- IBM and Together AI sign $240m multi-year cloud agreement
IBM and Together AI sign $240m multi-year cloud agreement verdict.co.uk
- ModelBest Begins Pre-IPO Tutoring as On-Device AI Competition Heats Up
ModelBest Begins Pre-IPO Tutoring as On-Device AI Competition Heats Up Caixin Global
- Chinese AI startup ModelBest starts pre-IPO process
In mainland China, pre-IPO tutoring is a mandatory stage before a company can formally file for an IPO.
- River AI secures $1.1bn investment for full-stack AI development
River AI secures $1.1bn investment for full-stack AI development verdict.co.uk
- SpaceXAI's Grok 4.6 matches OpenAI's best model and undercuts it on price
xAI's Grok 4.6 scores 61 points on the Artificial Analysis Intelligence Index, tying GPT-5.6 Sol and trailing only Anthropic's Claude Opus 5. On agentic tasks, it completes complex workflows in about 53 steps where Claude Opus 5 needs 103, at a price more than 60 percent lower. The article SpaceXAI's Grok 4.6 matches OpenAI's best model and undercuts it on price appeared first on The Decoder .
- Grok 4.6 returns SpaceXAI to the intelligence frontier and leads on cost efficiency
Grok 4.6 benchmark shows SpaceXAI leading in cost efficiency.
- Introducing Grok 4.6
New version of Grok with long-running agents and advanced interactive and visual capabilities.
- SpaceX Jumps: New Grok Model Released; Elon Musk Holds Companywide Meeting
SpaceX released its newest AI model, Grok 4.6, on Wednesday. The stock rose as much as 11%, rising above its IPO price. The post SpaceX Jumps: New Grok Model Released; Elon Musk Holds Companywide Meeting appeared first on Investor's Business Daily .
- Nvidia’s Nemotron 4 could become a 1-trillion-parameter AI model
Nvidia’s Nemotron 4 could become a 1-trillion-parameter AI model YourStory.com
- Linforge
Turn Anki flashcards into real English conversations
- Nvidia building 1-trillion-parameter Nemotron 4 to rival open AI models: Report
Nvidia building 1-trillion-parameter Nemotron 4 to rival open AI models: Report
- Nvidia's Nemotron 4 aims for one trillion parameters, a scale Chinese labs already surpassed
Nvidia is working on Nemotron 4, a new open-weight model designed to rival the world’s best freely available models. The article Nvidia's Nemotron 4 aims for one trillion parameters, a scale Chinese labs already surpassed appeared first on The Decoder .
- Nvidia’s New Open Model Is for Specific Use Cases
The model is small enough for enterprises to run on local devices and is targeted toward specific tasks.
🤖 ModelsAug 12, 2026https://aibusiness.com/agentic-ai/nvidia-s-new-open-model-for-specific-use-cases - Class Activation Mapping in Explainable Computer Vision: A Method-Centered Review of CNN, Transformer, and Foundation-Model-Era Visual Explanations
Class activation mapping (CAM) is one of the most widely used visual explanation families in explainable artificial intelligence. Its purpose is intuitive: it converts internal model evidence into a heatmap that highlights the image regions, convolutional channels, tokens, or patches that support a ...
- Structural Silence: When AI Infrastructure Fails Speakers of Underrepresented Languages
Artificial intelligence tools for education and language support are increasingly framed as scalable responses to access gaps in under-resourced communities. Yet the infrastructure underlying these tools, including training corpora, tokenization schemes, evaluation benchmarks, and deployment archite...
- Convergent Detour Hijacking: Task-Preserving Resource Amplification in Skill-Based LLM Agents
LLM agents increasingly rely on third-party skills, using natural-language descriptions for selection and instruction bodies for planning. This progressive-disclosure design exposes two sequential control points to untrusted publishers: a static skill may steer an otherwise correct task onto an unne...
- Diagram-MMU: A Multi-Modal Benchmark for Scientific Diagrams
Multimodal Large Language Models (MLLMs) have been growing the capability for scientific writing and collaboration. For example, OpenAI Prism is a free workspace for scientific writing and collaboration. One important feature in Prism is turning scientific diagrams directly into LaTeX TikZ code. In ...
- One Frozen Simulator Is Not Enough: Simulator Collapse in Multi-Agent RL
Multi-agent reinforcement learning for human-AI interaction typically relies on a single large language model to simulate user behavior. We show that this approach systematically fails to generalize, and trace the failure to simulator collapse: because the simulator LLM is mode-collapsed, an LLM pol...
- Domain-Aware Lightweight Spectral-Grouped Convolutions for Hyperspectral Fish Freshness Classification
Hyperspectral imaging (HSI) offers nondestructive assessment of fish freshness by detecting biochemical alterations across spectral bands. However, conventional deep learning approaches do not fully address the particular characteristics of HSI data, such as spectral dominance over spatial textures,...
- SCOUT: Unlocking Enhanced Spatial Reasoning via Structured Chain-of-Thought and Multi-Objective Process Reward
Existing Vision-Language Models (VLMs) exhibits a critical bottleneck in robust spatial reasoning. Recent reinforcement learning (RL) methods aim to close this gap with verifiable outcomes, yet they suffer from poor credit assignment across intermediate reasoning steps. Concurrently, structured reas...
- Learning-Based Behavior Planning for Automated Driving: Real-World Integration and Deployment
Recent research in machine and deep learning has shown the potential of learningbased motion planning approaches to improve the driving behavior of automated vehicles, especially in complex environments. However, their complex nature and lack of transparency can hinder explainability and trustworthi...
- NetlistBench: Evaluating LLM Reliability in SPICE Netlist Recognition and Manipulation
Large Language Models (LLMs) are increasingly used in circuit design workflows, yet their reliability on simulator-facing SPICE netlist recognition and manipulation remains poorly understood and is rarely separated from high-level design reasoning. Although netlists are textual, they encode structur...
- M-Net: Integrating Spectral Features and Physical Field Operators into Deep Learning for Medical Image Segmentation
Purpose: Deep learning-based medical image segmentation has achieved remarkable success, yet purely data-driven approaches often fail to exploit the rich mathematical structure inherent in medical images. We investigate whether explicit mathematical inductive biases, specifically matrix spectral ana...
- Machine Learning-Based Cyber Defense for Cloud Infrastructure: An Adaptive Deep Q-Network Architecture for Intelligent Intrusion Detection and Automated Threat Mitigation
With the increasing complexity of cyber assaults in cloud environments, adaptable security solutions are needed that can support real-time detection and autonomous response. In this paper, we propose a reinforcement learning-based dynamic cyber defense framework. We deploy a Deep Q-Network (DQN) to ...
- Who Thinks Best Depends on How Long You Let Them: Budget-Dependent Rankings in LLM Evaluation
Standard evaluation of large language models assumes stable model rankings across inference conditions. We challenge this assumption by varying the token generation budget, i.e., the maximum tokens a model may produce, across seven levels (64--4,096), evaluating four models on three reasoning benchm...
- Hybrid Gated Attention
Gated attention is an effective approach to mitigate attention sinks and enhance the representational capacity of attention. To further extend its effectiveness-efficiency Pareto frontier, we propose a Hybrid Gated Attention (HyGA) framework that contains three types of gating strategies. Specifical...
- A corpus-specific clinical RAG system matches or outperforms newer frontier LLMs on HealthBench
General-purpose large language models (LLMs) have recently been reported to match or exceed specialized clinical AI tools on medical benchmarks, but such comparisons draw on a narrow set of systems and on benchmarks developed largely in high-income settings. We evaluate VITA, a retrieval-augmented g...
- GUIDE: Governed Unified Intelligence for Document-to-Artifact Generation in Enterprise Settings
Enterprise guideline documents are heterogeneous and multimodal, combining narrative text, complex tables, and embedded images. Existing LLM and VLM systems face hallucinated content, table structure degradation, and lack governed workflows extending beyond extraction to validation and artifact gene...
- Do LLMs Take Care of Their Own? Similarity Signals Can Induce Cooperation
As LLM-based agents with user-instructed goals are becoming widely deployed, they increasingly encounter each other in strategic interactions, and face challenges of finding mutually beneficial outcomes. Prior literature has argued that cooperation problems such as the Prisoner's Dilemma are resolva...
- Ready Cohorts: Bounding GPU Opportunity and Avoiding Host Round Trips in LLM-Agent Control
LLM-agent services repeatedly execute small deterministic transitions between model and tool calls: route an outcome, update state, and emit the next effect. We ask when this control path exposes enough concurrent work for GPU execution, and what changes when a GPU-computed route decision remains on...
- QV-PIC: Query-Aware Visual Position-Independent Caching for Efficient RAG Serving
Retrieval-Augmented Generation (RAG) repeatedly prefills identical text chunks across queries, incurring redundant computations. Position-Independent Caching (PIC) mitigates it by reusing precomputed Key-Value (KV) across positions, but its efficiency is constrained by the large volume of text token...
- Confidence Calibration of Deep Learning Systems
In high-stakes applications, reliable confidence estimates are as important as the predictions themselves. Confidence calibration ensures that predicted probabilities reflect the likelihood of correctness, making it essential for safe deployment of deep learning models. However, existing methods typ...
- Experimental hypoxia to probe neuro-metabolic and vascular dysregulation in ME/CFS: a multimodal proof-of-concept MRI study
Background Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) is a poorly understood, debilitating multisystem condition. Converging evidence implicates impaired cellular bioenergetics, neuroinflammation and defective neurovascular coupling that may manifest as "virtual hypoxia" only under physiological stress. Methods We performed a single-session multimodal 3T MRI study combining brain volumetry, arterial spin labelling (ASL) and multivoxel proton magnetic resonance spectroscopy under normoxia and two controlled hypoxic challenges (oxygen saturation 87 {+/-} 3%) in 26 ME/CFS patients and 27 age- and sex-matched healthy controls. Results After intracranial-volume normalization, patients showed a reduced brainstem volume (1.46 0.14 vs. 1.55 {+/-} 0.18 % of eTIV; p = 0.013, FDR-p = 0.039), whereas deep grey matter and whole-brain parenchymal fraction did not differ between groups. Whole-brain cerebral blood flow (CBF) rose under hypoxia in both groups (controls +4.8 {+/-} 13.0%, patients +3.7 {+/-} 11.7%), with greater initial inter-individual variability in patients (patient-to-control variance ratio up to 6.94; FDR-p = 0.001). Thalamic lactate-to-creatine (Lac/tCr) ratios increased with hypoxia in controls (FDR-p = 0.028) but were already elevated at normoxia in patients (0.171 vs. 0.135; FDR-p = 0.021) and did not rise further (FDR-p = 0.38). In exploratory analyses, patients showed exaggerated inverse coupling between thalamic total N-acetylaspartate (tNAA/tCr) and white-matter CBF. Conclusions These findings provide in vivo evidence of impaired neuro-metabolic and vascular adaptive capacity in ME/CFS, supporting the virtual hypoxia hypothesis and highlighting candidate imaging markers for stratification that warrant validation.
- A Vision-Language Model for Coronary Angiography Interpretation and Clinical Decision Support
BACKGROUND: Coronary angiography remains the reference standard for diagnosing coronary artery disease and guiding revascularization, yet its interpretation requires expert integration of multi-view anatomy, lesion morphology and procedural context. Existing artificial intelligence approaches are largely task-specific, annotation-dependent and limited in capturing the semantic relationship between angiographic findings and interventional decision-making. Whether large-scale vision-language pretraining can enable transferable foundation-model representations for invasive coronary imaging remains unknown. METHODS We developed CAG-MIND, a domain-specific vision-language foundation model for coronary angiography, using 135,475 CAG examinations paired with procedural reports, comprising 812,850 angiographic videos from Zhongshan Hospital and Shanghai Geriatric Medical Center. Each case consisted of standardized six-view angiographic acquisitions paired with structured procedural semantics extracted from routine reports using a large language model-assisted pipeline. The model was pretrained by aligning multi-view angiographic representations with report-derived semantic embeddings through bidirectional contrastive learning. Performance was evaluated under zero-shot and supervised fine-tuning settings across 11 downstream tasks grouped into structural abnormality detection, atherosclerotic plaque assessment, and interventional decision prediction, using both an internal validation cohort and an independent external test cohort. RESULTS CAG-MIND demonstrated robust performance across all three task categories. In the zero-shot setting, the model achieved mean AUROCs of 0.686 in the internal validation cohort and 0.745 in the external test cohort, indicating transferable multimodal representations without task-specific supervision. Following supervised fine-tuning, the mean AUROC increased to 0.827 and 0.846, respectively, with excellent performance for coronary stenosis detection (AUROC 0.940 in both cohorts), balloon/stent prediction (0.900 and 0.907), and CABG recommendation (0.877 and 0.875). Compared with representative biomedical vision-language models and conventional image-based architectures, CAG-MIND consistently achieved superior performance in both zero-shot and supervised settings and remained superior to fully fine-tuned competing models when trained with only 10% of the labelled data. Grad-CAM visualization demonstrated anatomically plausible lesion-focused attention, supporting the interpretability of the learned representations. CONCLUSIONS CAG-MIND is, to our knowledge, the first large-scale vision-language foundation model for coronary angiography trained at more than 100,000-patient scale. By aligning standardized multi-view angiographic videos with report-derived procedural semantics, CAG-MIND enables robust zero-shot transfer, data-efficient fine-tuning and cross-center generalization. These findings support domain-aligned multimodal pretraining as a scalable foundation-model paradigm for invasive cardiovascular imaging and future cath-lab decision support.
- Predictors of Visual Acuity, Intraocular Pressure, and Pain in Neovascular Glaucoma: A Mixed-Effects Model and Machine Learning Cohort Analysis
Background/Aims: Neovascular glaucoma (NVG) is a severe, secondary glaucoma. This study aimed to identify factors associated with vision, intraocular pressure (IOP), and ocular pain outcomes. Methods: The cohort included all patients diagnosed with NVG during 2008-2024 at Helsinki University Hospital, Finland. Linear mixed-effects models used pre-specified covariates, whereas machine learning was given the full longitudinal data with biomicroscopic findings as an exploratory approach. Results: 626 patients were analysed. Worse baseline vision and a closed angle were associated with worse follow-up vision. Treatments were associated with lower IOP and less pain rather than better vision. Age, sex and comorbidity were largely not associated with the outcomes. Glaucoma drainage devices showed the greatest initial IOP reduction (-10.2 mmHg, 95% confidence interval, CI -11.9 to -8.6 mmHg), followed by transscleral cyclophotocoagulation (TSCPC, -4.7 mmHg, 95% CI -5.8 to -3.7 mmHg) and peripheral retinal cryotherapy (-2.2 mmHg, 95% CI -3.1 to -1.4 mmHg). TSCPC and cryotherapy were also associated with reduced pain (odds ratio 0.51 and 0.46). Pan-retinal photocoagulation and anti-VEGF showed smaller IOP reductions, with a pain reduction for pan-retinal photocoagulation only. Both methods agreed, and machine learning added no novel clinical findings. Conclusions: Vision in this cohort was largely set by the state of the eye at diagnosis. IOP control and pain relief therefore remain realistic goals even when sight cannot be saved. Peripheral retinal cryotherapy stood out, linked to both lower IOP and less pain, seldom reported in NVG. These associations from a large, unselected cohort identify treatments worth comparing prospectively.
- Gate-Before-Generate: A Dual-Layer Architecture for Output-Presence Routing in Chest X-ray Report Generation
CXR report generation may require a vision-language model (VLM) to produce both textual findings and spatial bounding boxes. Generative 4B-7B VLMs can emit non-empty outputs on normal images and empty outputs on abnormal images, motivating explicit structural routing. To evaluate whether a hard inference-time gate before a probabilistic VLM changes output-presence performance and to identify the mechanisms underlying paired STRUCT outcomes. We evaluated CXRxVLM v2, combining a frozen microsoft/rad-dino ViT-B/14 encoder with a 768[->]1 logistic probe (threshold 0.0557) and google/medgemma-4b-it with the pamessina/medgemma-4b-it-cure LoRA adapter. A seed=42 stratified cohort of 500 VinDr-CXR train-pool images (250 NORMAL, 250 ABNORMAL) was compared with Lingshu-7B A_baseline and D_fewshot configurations. Exact paired McNemar tests and stratum-level output-presence analyses were prespecified for the primary configurations; MedGemma 1.5 SigLIP was exploratory. CURE achieved STRUCT = 78.0% (390/500; Wilson 95% CI 74.2-81.4), versus 73.8% for Lingshu A_baseline and 74.2% for D_fewshot. Pairwise p-values were 0.0778, 0.1042, and 0.8642. The paired decomposition showed CURE ABNORMAL non-empty-output advantage of +13.6 percentage points versus Lingshu A (p = 0.0012; +14.0 points versus D, p = 0.0007), while Lingshu had higher NORMAL empty-output rates (+5.2 to +6.4 points; p = 0.0106 and p = 0.0004). The full pipeline used 8.87 GB VRAM and 4.92 s/image mean latency; 53% of records used a 25.7 ms warm gate-negative path after model loading. Equivalent overall STRUCT scores concealed two mechanistically different output regimes: CURE favored ABNORMAL non-empty outputs, whereas Lingshu favored NORMAL empty outputs. This paired decomposition, rather than the aggregate score alone, characterizes how hard-gated and probabilistic systems route output presence.
- Validation of a Novel Algorithm for Automated Detection and Quantification of Choroidal and Retinal Pulsation on Video Indocyanine Green Angiography
Purpose: To validate an automated pipeline to detect and quantify focal retinal and choroidal pulsation areas that are synchronous with the cardiac cycle in video indocyanine green angiography (ICGA). Design: Retrospective, observational, hypothesis-generating validation study Subjects, Participants: Consecutive patients with a diagnosis of central serous chorioretinopathy (CSCR) in one or both eyes. Methods: Videos were acquired on Heidelberg HRA+OCT. The pipeline consisted of three steps: signal extraction, foci detection, and quantification. After registration of the constituent frames, each pixel's intensity signal was analyzed at the presumed cardiac frequency (tested from a sample of three detectable frequencies). A synchrony score combining local phase coherence with oscillation amplitude was then derived and computed using a standard deviation ({sigma}) above each video's background oscillation value. Two masked graders marked the retinal and choroidal pulsation areas twice. We compared detection of the pulsation areas against grader consensus using a receiver operating characteristic curve (using multiple grid sizes to divide the scan area) and, separately, using a signal-based area-reduction method to obtain an optimum {sigma} value. Main Outcome Measures: Agreement between the automated algorithm and human graders in detection of pulsation foci, and the optimum threshold multiplier ({sigma}). Results: We studied 20 ICGA videos from 20 eyes. At the 16-pixel grid size, the pipeline achieved a mean area under the curve (AUC) of 0.914, sensitivity of 0.86, and specificity of 0.80. Grader agreement improved with larger grid size, reaching substantial-to-strong levels for choroidal annotations. The two independent validation methods demonstrated similar {sigma} values that differed by 0.62{sigma}, supporting {sigma}=4.0 as the optimum value. Conclusions: We report the first automated method to quantify retinal and choroidal vascular pulsation on video ICGA. It measures pixels that oscillate over time with the presumed cardiac cycle and works reliably at the spatial scale (grid level) where experts agree. Pulsatile hemodynamics may add a new vascular biomarker for glaucoma, diabetes, hypertension, and pachychoroid diseases.
- The Dietary Approaches to Stop Hypertension (DASH) diet score and its association with the Risk of Kidney Function Decline and Mortality among Veterans in the Million Veteran Program
Background: The dietary approaches to stop hypertension (DASH) diet is often recommended to patients with chronic kidney disease, although evidence regarding its efficacy in this population is limited. Our study tested the hypothesis that increased adherence to the dietary approaches to stop hypertension (DASH) diet score would be associated with longer time to kidney function decline among Veterans. Methods: We conducted a retrospective cohort study of 251,921 Veterans enrolled in the Million Veteran Program (MVP). The DASH diet score was calculated from the food frequency questionnaire and categorized into tertiles. The primary outcome was a composite of: Kidney event or death, where a kidney event was defined as a sustained 40% decline in estimated glomerular filtration rate (eGFR) or end-stage kidney disease (ESKD). Cox regression models compared the hazard for both outcomes by DASH score tertiles. We examined modification by ancestry, sex and other clinical characteristics Results: The median age was 67 years and 90% of Veterans were men. There were 59,269 (23.5%) who experienced the primary composite outcome, during the maximum follow-up of 10 years (median 6.1 years). Crude incidence rates for the kidney event and death outcome were 43.3, 40.4, and 36.8 per 1000 person-years of DASH score by tertiles. DASH score was associated with a lower hazard ratio (HR) for the primary composite outcome; third vs first tertile 0.81 (95% Confidence Interval (CI) 0.80 - 0.83) and second vs first tertile HR 0.90 [95% CI 0.88 - 0.92]. In subgroup analysis for individuals of African ancestry, Admixed American, and Females, only the third tertile of the DASH score was associated with a statistically significant reduction in composite outcome. Conclusion: Beneficial associations of the DASH diet were observed across subgroups. Future research is needed to understand gene and environmental factors that influence the observed subgroup differences
- Graph-Structured Rubrics: Compiling Rubrics into Typed Evaluation Graphs for LLM Judges
Rubric-based evaluators commonly treat rubrics as prompt context or flat criteria: they specify what to judge but leave criterion composition implicit, even when natural-language rules state it. We introduce Graph-Structured Rubrics (GSR), which compiles a rubric into a response-independent typed ev...
- Faithful, Sufficient and Understandable: Rethinking Graph Counterfactual Explanations via Discrete Diffusion Inversion
Graph Neural Networks (GNNs) achieve strong predictive performance on graph-structured data across domains such as chemistry, biology, and network analysis, yet they provide no intrinsic explanation of their predictions. This limits their adoption in high-stakes and safety-critical settings. Counter...
- Learning Loco-Manipulation From SMPC Demonstrations With Sparse Offline-to-Online RL
Integrating locomotion and manipulation is essential for robot autonomy, but scaling standard Reinforcement Learning (RL) to complex tasks is severely bottlenecked by the slow, manual process of dense reward shaping. To bypass this limitation, we leverage Sample-based Model Predictive Control (SMPC)...
- From Safety Documentation to Safety Knowledge Support: An Evidence-Grounded LLM Framework for Medical Devices
Medical devices are becoming more software-intensive, connected, and AI-enabled. Their development requires risk-management evidence aligned with ISO 14971 and, for software, IEC 62304. This evidence must be kept consistent across requirements, design decisions, software changes, verification result...
- Dual-Model Sentiment Analysis of Consumer Reviews in the Retail Coffee Sector Using Machine Learning and Deep Learning Approaches
Consumer reviews play an important role in shaping brand perception and business strategies, particularly in service-driven industries such as retail coffee. This study presents a comparative sentiment analysis framework for Starbucks customer reviews using classical machine learning and deep learni...
- RealisticTritonBench: A Benchmark for Triton-Kernel Generation in Real-World AI Frameworks
In modern AI frameworks, GPU kernels are key to overall system performance. Combining usability, portability, and near-handwritten CUDA performance, Triton is widely adopted for implementing GPU kernels. Recent advances show the potential of large language models (LLMs) to automatically generate Tri...
- DreamFly: Causal Memory and Receding-Horizon Diffusion Planning for Aerial Vision-Language Navigation
Aerial vision-language navigation (VLN) requires an embodied agent to integrate visual evidence over time, plan future actions, and determine when it has reached a navigation goal under partial observability. Although recent VLA models offer a promising perception-to-action paradigm, adapting them t...
- AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses
Recent work on distillation transfers the capabilities of large models to smaller ones often by updating the latter's parameters, through teacher forcing, on-policy distillation, and related training-time methods. In this paper, we ask whether such transfer can instead occur at test time. We study s...
- Redistribution-based Cost Inference Improves Sparse Safe Offline RL
Safe offline RL typically assumes access to dense per-step cost annotations, but in practice supervisors provide only trajectory-level stop-feedback: a binary signal at the first unsafe transition, with no per-step attribution. We frame this as a temporal credit assignment problem and propose the Re...