Tag
ABot-Recon is an open-source real-time streaming 3D reconstruction model from Amap CV Lab that uses KV cache context and per-frame estimation to generate updatable 3D point cloud maps, achieving 24.45 FPS on H100 GPU and deployable on edge devices like AX650N NPU.
The paper presents AHL Studio, a tool using agentic heuristic learning to create executable, inspectable policies for human activity recognition without traditional training, targeting edge deployment.
This paper introduces a carbon-aware routing framework for function-calling LLMs in edge-cloud systems, reducing operational carbon emissions by an average of 4× while maintaining cloud-level accuracy.
smolbenchmark is a new resource that benchmarks small AI models on consumer hardware devices, providing metrics like speed, efficiency, and thermals to help users choose the best model for their specific hardware.
Desert Ant Labs offers small, specialized AI models for speech, text, and vision that run offline on devices, with an SDK for easy integration and no per-use costs.
MiniCPM5-2B is a 2B-parameter open-source language model that achieves high intelligence density for edge deployment and ranks highly in several AI performance benchmarks.
AirCaps Audio Research Lab launches with streaming speech and audio models designed for complex real-world environments, claiming superior performance on edge hardware compared to leading cloud models.
A tweet speculates that Apple, Qualcomm, and Google could become leading AI companies by focusing on portable, edge AI devices, referencing Google Gemma running on NVIDIA Jetson.
Maxime Rivest predicts Qualcomm could replicate NVIDIA's success in edge AI, citing recent acquisitions and advancements in mobile AI training efficiency compared to historical NVIDIA GPUs.
This paper introduces a PyTorch-based framework for pruning binarized neural networks and proposes a novel global weighting pruning method that achieves superior accuracy and pruning rates, enabling efficient deployment on edge hardware like FPGAs.
This paper proposes a SAREF-compliant ontology for representing and orchestrating distributed AI workflows across heterogeneous edge, fog, and cloud environments to improve semantic interoperability and resource-aware deployment.
This paper identifies an Accuracy-Efficiency Paradox in on-device energy forecasting where high-precision models can cause net energy loss due to inference energy consumption and battery aging, and proposes a Total Cost of Ownership framework to minimize this loss.
audio.cpp version 0.7 introduces a new Arena UI for comparing audio models locally, expands to 62 model families with over 85 variants, and supports deployment on edge hardware like NVIDIA Jetson Orin.
The article introduces Polign DB, a typed, stateless database designed for agent memory, optimized for edge deployment and cost efficiency. It demonstrates a demo showing how it can manage structured memory for AI agents with minimal resource usage.
This paper proposes a Sparse-Activation-ReLU (SAR) layer for low-latency, energy-efficient virtual sensing, achieving significant improvements in latency-error-energy metrics and reducing errors through synthetic knowledge distillation.
Arduino VENTUNO Q is a new AI-powered development board combining Qualcomm and STM32 processors to run local LLMs and AI models for physical computing, offering low-latency edge AI with real-time control capabilities.
FL-MAESTRO introduces a multi-agent LLM orchestrator to jointly optimize communication topology, resource allocation, and aggregation rules in federated learning, significantly reducing wasted energy and improving efficiency on resource-constrained edge devices.
FreeToken enables running large AI models like Qwen3.6 35B on limited hardware such as an 8GB GPU by offloading to CPU and RAM, providing a cost-effective solution for edge computing.
The article breaks down the robotics stack into six key layers, from materials and power to deployment, emphasizing the shifting constraints in the industry as it evolves.
This paper proposes telemetry-informed adaptive compression for edge-based RAG systems, showing experimental evidence that intermediate compression can reduce GPU energy by up to 53.2% with negligible quality loss.