Awesome research works for on-device AI systems

A curated list of research works on efficient on-device AI systems, methods, and applications for mobile and edge devices.

By Topic

Hardware-aware Attention Acceleration Methods

• [MLSys 2025] MAS-Attention: Memory-Aware Stream Processing for Attention Acceleration on Resource-Constrained Edge Devices [paper]
• [MLSys 2025] Efficient LLM Inference using Dynamic Input Pruning and Cache-Aware Masking [paper]
• [NeurIPS 2022] FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness [paper]

On-device LLM Inference Systems

• [arXiv 2025] HeteroLLM: Accelerating Large Language Model Inference on Mobile SoCs with Heterogeneous AI Accelerators [paper]
• [ASPLOS 2025] Fast On-device LLM Inference with NPUs [paper] [code]
• [arXiv 2024] PowerInfer-2: Fast Large Language Model Inference on a Smartphone [paper]
• [MobiCom 2024] MELTing point: Mobile Evaluation of Language Transformers [paper] [code]
• [MobiCom 2024] Mobile Foundation Model as Firmware [paper] [code]

Inference Acceleration Systems with Heterogeneous Computing Processors (e.g., CPU, GPU, NPU, etc.)

• [MobiSys 2024] Pantheon: Preemptible Multi-DNN Inference on Mobile Edge GPUs [paper]
• [MobiCom 2024] Perceptual-Centric Image Super-Resolution using Heterogeneous Processors on Mobile Devices [paper]
• [Sensys 2023] Miriam: Exploiting Elastic Kernels for Real-time Multi-DNN Inference on Edge GPU [paper]
• [MobiSys 2023] NN-Stretch: Automatic Neural Network Branching for Parallel Inference on Heterogeneous Multi-Processors [paper]
• [ATC 2023] Decentralized Application-Level Adaptive Scheduling for Multi-Instance DNNs on Open Mobile Devices [paper]
• [IPSN 2023] PointSplit: Towards On-device 3D Object Detection with Heterogeneous Low-power Accelerators [paper]
• [SenSys 2022] BlastNet: Exploiting Duo-Blocks for Cross-Processor Real-Time DNN Inference [paper]
• [MobiSys 2022] Band: Coordinated Multi-DNN Inference on Heterogeneous Mobile Processors [paper]
• [MobiSys 2022] CoDL: efficient CPU-GPU co-execution for deep learning inference on mobile devices [paper]

Adaptive Inference Systems for Optimized Resource Utilization

• [RTSS 2024] FLEX: Adaptive Task Batch Scheduling with Elastic Fusion in Multi-Modal Multi-View Machine Perception [paper]
• [MobiCom 2024] Panopticus: Omnidirectional 3D Object Detection on Resource-constrained Edge Devices [paper]
• [MobiSys 2023] OmniLive: Super-Resolution Enhanced 360° Video Live Streaming for Mobile Devices [paper]
• [MobiSys 2023] HarvNet: Resource-Optimized Operation of Multi-Exit Deep Neural Networks on Energy Harvesting Devices [paper]
• [MobiCom 2022] NeuLens: Spatial-based Dynamic Acceleration of Convolutional Neural Networks on Edge [paper]
• [MobiCom 2021] Flexible high-resolution object detection on edge devices with tunable latency [paper]

On-device Training, Model Adaptation Systems

• [ASPLOS 2025] Nazar: Monitoring and Adapting ML Models on Mobile Devices
• [SenSys 2024] AdaShadow: Responsive Test-time Model Adaptation in Non-stationary Mobile Environments [paper]
• [SenSys 2023] EdgeFM: Leveraging Foundation Model for Open-set Learning on the Edge [paper]
• [MobiCom 2023] Cost-effective On-device Continual Learning over Memory Hierarchy with Miro [paper]
• [MobiCom 2023] AdaptiveNet: Post-deployment Neural Architecture Adaptation for Diverse Edge Environments [paper]
• [MobiSys 2023] ElasticTrainer: Speeding Up On-Device Training with Runtime Elastic Tensor Selection [paper]
• [SenSys 2023] On-NAS: On-Device Neural Architecture Search on Memory-Constrained Intelligent Embedded Systems [paper]
• [MobiCom 2022] Mandheling: mixed-precision on-device DNN training with DSP offloading [paper]
• [MobiSys 2022] Memory-efficient DNN training on mobile devices [paper]

Profilers

• [SenSys 2023] nnPerf: Demystifying DNN Runtime Inference Latency on Mobile Platforms [paper]
• [MobiSys 2021] nn-Meter: towards accurate latency prediction of deep-learning model inference on diverse edge devices [paper]

By Conference (2025~)

MLSys 2025

• MAS-Attention: Memory-Aware Stream Processing for Attention Acceleration on Resource-Constrained Edge Devices [paper]
• Efficient LLM Inference using Dynamic Input Pruning and Cache-Aware Masking [paper]
• TurboAttention: Efficient attention approximation for High Throughputs LLMs [paper]
• SampleAttention: Near-Lossless Acceleration of Long Context LLM Inference with Adaptive Structured Sparse Attention [paper]
• LeanAttention: Hardware-Aware Scalable Attention Mechanism for the Decode-Phase of Transformers [paper]

ASPLOS 2025

• Fast On-device LLM Inference with NPUs [paper] [code]
• Energy-aware Scheduling and Input Buffer Overflow Prevention for Energy-harvesting Systems
• Generalizing Reuse Patterns for Efficient DNN on Microcontrollers
• Nazar: Monitoring and Adapting ML Models on Mobile Devices

EuroSys 2025

• Flex: Fast, Accurate DNN Inference on Low-Cost Edges Using Heterogeneous Accelerator Execution [paper]
• T-MAC: CPU Renaissance via Table Lookup for Low-Bit LLM Deployment on Edge [paper]

SOSP 2025

MobiSys 2025

MobiCom 2025

Preprint 2025

• HeteroLLM: Accelerating Large Language Model Inference on Mobile SoCs with Heterogeneous AI Accelerators [paper]