Papers for

edge ai developers

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

Event-driven graph neural network processor achieves ultra-low latency for edge AI

A 25-$μ$s/inf Event-driven Graph Neural Network Processor with Spatiotemporal Caching and Spline Convolution for Ultra-low-latency AI at the Edge

Abstract: Dynamic-vision-sensor (DVS) cameras generate events on a per-pixel basis with a $μ$s-level temporal resolution, calling for new algorithm-hardware co-design approaches compared to standard frame-based vision. While event-driven graph neural networks (EV-GNNs) emerge as a promising algorithmic solution, they raise new HW challenges by mixing dense-regular compute operations and sparse-irregular memory accesses. We present ETHEREAL, the first EV-GNN accelerator that scales to 640$\times$480 resolutions, thanks to a neighbor-parallel spline convolution engine and a 2D/3D-split memory hierarchy with a novel region-of-interest spatiotemporal caching mechanism. Measurement results demonstrate end-to-end inference with 25.6$μ$s latency and 1.7$μ$J energy per event on state-of-the-art workloads

Mon 14 SeptComputer Vision and Pattern RecognitionHardware Architecture
The gist
Dynamic vision sensors create data events very quickly and individually for each pixel, which challenges standard computer vision methods. The authors developed ETHEREAL, a specialized chip that processes these rapid events efficiently by combining new memory and computation techniques. This design enables very fast analysis of visual data, with results available in about 25 microseconds and very low energy use. Their work helps machines use event-based cameras more effectively in real-time applications.
Open 2609.15241v1

Flexible spiking neural network chip speeds up and saves energy in vision tasks

FlexSpIM: An Event-Based Digital Compute-In-Memory Accelerator with Flexible Operand Resolution and Layer-Wise Hybrid Stationarity

Abstract: Compute-in-memory (CIM) accelerators for spiking neural networks (SNNs) offer a promising solution for achieving $μ$s-level inference latency and ultra-low energy in edge vision applications. However, their limited flexibility at both circuit and system levels restricts their deployment across diverse workloads. This work introduces FlexSpIM, a digital CIM architecture supporting arbitrary operand resolution and shape within a unified storage for weights and neuron states (i.e., membrane potentials). These circuit-level capabilities enable a layer-level hybrid weight- and output-stationary dataflow, maximizing operand reuse and reducing costly on- and off-chip data movement during SNN execution. Measurement results from a fabricated FlexSpIM prototype in 40-nm CMOS demonstrate competitive 1-bit-normalized energy efficiency and higher throughput compared with prior fixed-precision digital CIM-based SNN accelerators, while providing bitwise resolution reconfiguration. Evaluated on the IBM DVS gesture dataset, FlexSpIM achieves 95.8% accuracy while enabling up to 45% energy and 52% latency reductions in large-scale systems compared with fixed stationarity approaches.

Tue 8 SeptHardware Architecture
The gist
Spiking neural networks (SNNs) are a type of artificial brain model that can quickly and efficiently process information for things like vision. The authors created FlexSpIM, a special chip that performs calculations right where data is stored, allowing it to work faster and use less energy. This chip can change how precise its calculations are and how it handles data during processing, making it adaptable to different tasks. Tests showed that FlexSpIM is more efficient and faster than earlier similar chips, maintaining high accuracy on gesture recognition.
Open 2609.08446v1

Unified system improves speed and cost of large AI model use

Unified AI Gateway: A Framework for Joint Model Routing and KV Cache Management

Abstract: Large language model (LLM) inference increasingly spans models that differ in size, capability, price, and provider. This shift creates two costs for developers. One is the integration cost of choosing among and switching between many models. The other is the inference cost of rebuilding a KV cache when it is unavailable or incompatible with the selected model. We define and analyze the Unified AI Gateway as a system setting for an edge-deployed AI traffic hub. It coordinates model routing, KV cache management, and compute placement across end devices, edge resources, and cloud model services. At request time, the gateway jointly selects a target model, an execution site, and a KV cache action under task-quality, latency, cost, and resource constraints. In parallel, background cache-management actions optimize KV cache placement, replication, retrieval, and lifecycle decisions for subsequent requests. We synthesize existing evidence on KV cache reuse, compression, cross-model mapping, distributed storage, and transfer, and discuss the remaining challenges of integrating these capabilities into one system. Across eight typical workload profiles, our workload-level analytical simulation reports TTFT speedups of 1.25$\times$--13.28$\times$ and input-cost benefits of 1.20$\times$--6.16$\times$.

Mon 7 SeptDistributed, Parallel, and Cluster Computing
The gist
Using large AI language models can be tricky because there are many models with different costs and abilities, and switching between them is complicated. The authors propose a system called the Unified AI Gateway that smartly chooses which model to use, where to run it, and how to manage memory caches to save time and money. This system works across devices, edge servers, and cloud services. Their tests show it can make AI responses faster and cheaper under various workloads.
Open 2609.06940v1