The Archive

Neural emulator with FiLM conditioning corrects marine biogeochemistry model bias using satellite data.

Quantum-inspired CNN generates pseudo-labels for land-cover classification from SAR and multispectral satellite data.

SymKL unifies cross-modality signal degradation measurement using implicit neural representations and Fisher information geometry.

UAV-based traffic control via bird's-eye-view imagery and neural networks; domain-specific application without general-purpose builder utility.

Neural network amortizes branch-and-bound upper bounds for probabilistic graphical model inference; theoretical optimization without practical tooling.

Tutorial covering neural codecs, autoregressive models, diffusion, flow matching, and preference optimization for music audio generation.

Convex-SKM-Net: GPU-optimized constraint satisfaction layer for learning-based decision-making via dual PGD.

Heterogeneous data distribution in federated learning improves neuroimaging model performance on Alzheimer's disease classification.

Structure-driven adversarial attack framework for GNNs using spectral graph properties; separates geometry from learned weights.

Dexterous manipulation with multi-fingered hands is the next frontier of robotics, yet learned hand skills remain limited to simple grasps or basic repeated motions, such as rotating an object. As robots attempt more demanding in-hand maneuvers, safety, such as not dropping the object, becomes critical, yet it is rarely measured. We study in-hand object pose reaching, a task that demands diverse and difficult grasps: moving an object to almost any position and orientation within reach of the fingers, at any wrist orientation, through long sequences of consecutive goals. We present a recipe fo...

CrossProfile benchmark reveals how LLM group memory heuristics (rare vs. relevant claims) fail on hidden-profile collaborative decision tasks.

Pneuma: public benchmark for aircraft predictive maintenance with 920M+ multivariate time steps, 9 fault scenarios, aligned to maintenance logs.

NVIDIA RTX Spark is a new Windows PC platform for AI development, creation, and gaming in slim laptops and compact desktops. Powered by an NVIDIA Blackwell RTX... NVIDIA RTX Spark is a new Windows PC platform for AI development, creation, and gaming in slim laptops and compact desktops. Powered by an NVIDIA Blackwell RTX GPU and NVIDIA Grace CPU, it combines up to 128 GB of unified memory with the full NVIDIA CUDA stack and RTX technologies including ray tracing and DLSS 5. This post covers the following technologies for game developers building on RTX… Source

Structure-encoded problems show that neural networks cannot learn symbolic structure efficiently without explicit domain knowledge, defying scaling assumptions.

Neural mechanism design for online resource allocation balances efficiency with strategyproofness via deep RL instead of linear rules.

Task Advancement Memory for GUI agents: tracks verified task progress to enable reusable operations across multi-app workflows.

Energy-efficient Spiking Neural Network accelerators for mmWave sensing and activity recognition; hardware-focused, not frontier AI modeling.

Physics-informed neural networks for quickest detection in high-dimensional diffusions; mathematical methods paper with limited AI architecture relevance.

NEXUS framework combines neural skills with interpretable symbolic meta-policies in hierarchical RL for transparency and performance.

ExpectileGCN proposes adaptive expectile-based aggregation for graph neural networks on heterophilic graphs.

Anthropic demonstrates Claude's capability to accelerate drug discovery workflows, showing practical application of AI in pharmaceutical R&D.

Theoretical analysis of how intrinsic noise in RNNs affects representation learning dynamics and convergence.

DECE benchmark: unified arena evaluating LLM agents across 10 Eurogames with shared strategic competition and resource constraints.

Independent third-party safety evaluation of DeepSeek V4.1 Flash-1017 for loss-of-control and manipulation risks following EU AI Code of Practice.

At the 2026 Open Compute Project (OCP) Global Summit, Meta shared the networking innovations that are directly supporting our latest AI models, including Muse Wonder and Muse Spark, and our gigawatt-scale Prometheus cluster. Highlights IcePack & IceCube: Meta’s first liquid-cooled network rack with 102.4T switch blades (128×800G, Broadcom Tomahawk 6C and NVIDIA Spectrum-6 variants), running [...] Read More... The post Meta Networking at the 2026 OCP Global Summit: Open Innovation for Scale-Up, Scale-Out, and Scale-Across AI Networks appeared first on Engineering at Meta.

Google releases "Polar Peaks" AI Quests educational lesson combining neuroscience and AI literacy.

Active learning framework using Transport Neural Tangent Kernel for automated experimental design with conditional generative models.

Neuroevolution framework TRON transfers protein-protein interaction network topologies to improve neural architecture search.

Alibaba and Nvidia executives pitch Europe on open-source AI models and industrial automation applications.

Introduces FlowGenesis: dynamic neural architecture adaptation via neuron birth and pruning guided by representation flux.