هذه الصفحة متاحة أيضًا باللغة العربية.
التبديل إلى العربيةEsta página también está disponible en español.
Cambiar a españolCette page est également disponible en français.
Passer en françaisBu sayfa Türkçe olarak da mevcut.
Türkçe sürüme geçArtificial intelligence research, applications, data infrastructure and verified corporate announcements. Page 4.
122 stories
BusinessEvent / data period:
The authors compare how output representation affects language-model generation of register-transfer-level hardware code. Their bounded task is generating SECDED error-correcting memory wrappers under formally checkable correctness conditions.
AgricultureEvent / data period:
The four young women developed a platform intended to track nurseries in one system and forecast demand and production.
BusinessEvent / data period:
A preprint introduces SpecKCast, a feature-forecasting method intended to reduce repeated transformer evaluations in diffusion models. It uses a spectral Kalman approach to predict features at skipped denoising steps, extending work on caching intermediate features.
BusinessEvent / data period:
Researchers present a method for three-dimensional thermal characterization of advanced 3D integrated circuits. The workflow extracts thermal properties by layer and builds a location-specific database to examine how heat moves through a stack.
BusinessEvent / data period:
A preprint presents LENS, a system for camera localization and neural scene rendering on an AMD KR260 FPGA and ARM platform. The work adapts a scene-coordinate and Gaussian-splatting pipeline to an edge device, with the goal of operating without a desktop GPU.
BusinessEvent / data period:
A preprint describes a workflow for mapping a small neural simulation of Caenorhabditis elegans to Intel’s Loihi-2 neuromorphic hardware through NeuroML2Loihi. The study addresses the cost of repeatedly updating neural and synaptic states in software.
BusinessEvent / data period:
Researchers introduce PACT, a system for LoRA fine-tuning on consumer GPUs that uses native NVFP4 matrix operations for frozen projections. The paper also addresses memory spent on activations and the cost of quantizing and dequantizing data.
BusinessEvent / data period:
EdgeKernelBench studies whether large language models can generate kernels for varied edge-computing hardware. The preprint introduces a common intermediate representation and a hierarchical benchmark that connects workloads to different backends.
BusinessEvent / data period:
A preprint introduces Koi, a cluster planner for large language model inference across heterogeneous multi-cloud fleets. It considers accelerator choice, parallelization, serving engines, replica counts and changing workloads as interdependent operating decisions.
BusinessEvent / data period:
A research preprint introduces Muse Glimmer, an open-source 30-billion-parameter causal language model with a perception encoder for agent tasks on consumer hardware. The authors describe text and image input, multi-step reasoning, tool use and failure recovery.
BusinessEvent / data period:
The preprint examines whether smartphone neural processing units can help fine-tune language models locally with low-rank adaptation. It focuses on the repeated operations involving frozen model weights that remain costly even when only a small number of parameters are trained.
BusinessEvent / data period:
The authors examine whether the matrix-vector processor inside a commodity EFR32MG24 microcontroller can accelerate sparse spiking-neural-network inference beyond an optimized Cortex-M33 CPU. Their toolchain maps snnTorch models to C and reorders neurons to improve accelerator use.
A multilingual perspective on the global economy.
