Neural Processing Units

Follow this topic Ask about this topic

Momentum

7 papers in the last four weeks, against 1 the four weeks before. 0.1% of all new papers.

Jul 13Week of Sep 28

Latest papers 31

All topics

From Experience to Expertise: Adoption-Aware Memory Learning for Data-Scarce NPU Kernel Synthesis
Sep 28, 2026Longxiao Fan, Tao Zhang, Han Yan +5Neural Processing Units Cross-Task Knowledge Transfer
Fudan University · University of Science and Technology of China · Huawei Technologies Ltd.
Dynamic Flow, Static Graph: KV Cache Reuse for Efficient LLM Serving on Mobile NPUs
Sep 28, 2026Zhengxiang Huang, Shengheng Chen, Chaoyue Niu +6Kv-Cache Management Neural Processing Units
Shanghai Jiao Tong University Shanghai, China · Alibaba Group Hangzhou, China · Alibaba Group Hangzhou, Zhejiang, China
COGNIT-Guard: Calibrated Standalone Direct-Decision Guardrails with Heterogeneous CPU-NPU Confidence Cascading under Explicit Latency and False-Positive Constraints
Sep 27, 2026Hao ChenStreaming Guardrails Large Language Model Safety
School of Cyber Security, Guangdong Polytechnic Normal University, Guangzhou, China
Does per-frame early exit pay? A compute-matched study of dynamic depth for on-device speech enhancement
Sep 24, 2026Clément Laroche, Riccardo MicciniSpeech Enhancement Neural Processing Units
GN A/S, Denmark · Technical University of Denmark, Denmark
Beyond Model Size: Redesigning LiSenNet for embedded speech enhancement
Sep 24, 2026Clément Laroche, Rasmus Kongsgaard OlssonSpeech Enhancement Neural Processing Units
GN A/S, Audio Research, Lauptrupbjerg 7, 2750, Denmark
NPBoost: Neural Processes with Gradient-Boosted Fixed Effects
Sep 23, 2026Andrea Nava, Ken Rölli, Armin Begic +1Meta-Learning Neural Processing Units
Seminar for Statistics, ETH Zurich, Switzerland · Lucerne University of Applied Sciences and Arts, Switzerland
RISC-V and machine learning: a survey
Sep 17, 2026Shriman Keshri, Apparna Singh, Chinmaya Kumar Palo +2Risc-V Neural Processing Units
National Institute of Science Education and Research, Bhubaneswar, Odisha, India · Work conducted at NISER · Sri Sri University, Cuttack, Odisha, India +2
DynaNDE: Dynamic Near-Data Expert Scheduling for Batched MoE Inference
Aug 31, 2026Xiaoyang Lu, Belthangady Akash Vi Narayana Pai, Xian-He SunMixture-Of-Expert Inference Neural Processing Units
Illinois Institute of Technology Chicago, Illinois, USA
AgenticCANN: Automated Ascend C Operator Generation via Knowledge-Augmented Agentic Evolution
Jul 29, 2026Junhao Qiu, Zidong Wang, Yansong Sun +3Neural Processing Units Hardware Efficiency
Department of Computer Science, City University of Hong Kong · Hong Kong Yuwin Technology Co., Ltd
QScheduler: Adaptive Gradient Sampling for Zeroth-Order On-Device Training on INT8 NPUs
Jul 21, 2026Victor Felipe Domingues Do Amaral, Pierre Demaj, Erwan Libessart +3Zeroth-Order Optimization Neural Processing Units
GeePs · STMicroelectronics, France · GeePs, CNRS, CentraleSup´elec, Universit´e Paris-Saclay, Sorbonne Universit´e, France
Mitigating Compiler Fusion-Induced Power Bursts in Mobile NPU Inference as the Battery Depletes
Jul 17, 2026Ryoga Yuzawa, Masayoshi TomizukaNeural Processing Units Voltage
UC Berkeley · Sony · Visiting researcher at University of California, Berkeley.
STEEL: Sparsity-Aware Fused Attention for Energy-Efficient Long-Sequence Inference on AMD's XDNA NPU
Jul 10, 2026Victor J. B. Jung, Gagandeep Singh, Joseph Melber +3Neural Processing Units Taxonomy-Driven Dataflow Model
Integrated Systems Laboratory (IIS), ETH Zürich, Switzerland. · AMD Research and Advanced Development (RAD). · Department of Electrical, Electronic and Information Engineering (DEI), University of Bologna, Italy.
Is Your NPU Ready for LLMs? Dissecting the Hidden Efficiency Bottlenecks in Mobile LLM Inference
Jul 6, 2026Guanyu Cai, Ruiming Tian, Lang Yang +4Neural Processing Units LLM Inference Optimization
Tsinghua University · Beijing Jiaotong University
AI-RAN on NPUs: Baseband Processing Without Baseband Chips
Jul 5, 2026Shilong Zhang, Luping Xiang, Jienan Chen +1Neural Processing Units Open Radio Access Networks
State Key Laboratory of Novel Software Technology, Nanjing University, Nanjing 210008, China · School of Intelligent Software and Engineering, Nanjing University (Suzhou Campus), Suzhou 215163, China · National Key Laboratory of Science and Technology on Communications, University of Electronic Science and Technology of China, Chengdu 611731, China
Hawk: Harnessing Hardware-Aware Knowledge for High-Performance NPU Kernel Generation
Jul 2, 2026Junyi Wen, Ruiyan Zhuang, Yongjia Xu +7Graphics Processing Unit Kernels Neural Processing Units
Sun Yat-sen University, Zhuhai, China · Greater Bay Area National Technology Innovation Center, Guangzhou, China · Peng Cheng Laboratory, Shenzhen, China
NPUsper: Eliminating Redundant Computation for Real-Time Whisper on Mobile NPUs
Jul 1, 2026Sihyeon Lee, Hojeong Lee, Sungwon Woo +3Whisper-Large-V3 Neural Processing Units
Korea University · University of Wisconsin–Madison
An Empirical Study of OpenPangu Quantization on Ascend NPUs
Jun 19, 2026Tong Shi, Jiacheng Wang, Hui Xie +4Post-Training Quantization Neural Processing Units
State Key Laboratory of Complex & Critical Software Environment, Beihang University · School of Artificial Intelligence, Beihang University · School of Computer Science and Engineering, Beihang University
KATANA: A Fast, Low-Power Mapping of Kalman Filters onto Edge NPUs for Real-Time Tracking
Jun 12, 2026Bodhisatwa Kundu, Anish Rooj, Sumit Saha +4Extended Kalman Filter Neural Processing Units
1Jadavpur University, West Bengal, India · 2Purdue University, IN, USA · 3Intel Corporation, CA, USA
Efficient On-Device Diffusion LLM Inference with Mobile NPU
Jun 11, 2026Tuowei Wang, Yanfan Sun, Ju RenNeural Processing Units LLM Inference Optimization
Beihang University China · Tsinghua University China
TileFuse: A Fused Mixed-Precision Kernel Library for Efficient Quantized LLM Inference on AMD NPUs
Jun 9, 2026Wesley Pang, Gregory Hyegang Jun, Feiyang Liu +1Neural Processing Units Matrix Multiplication
University of Illinois Urbana-Champaign Department of Electrical and Computer Engineering USA
Energy-Efficient On-Device RAG on a Mobile NPU: System Design and Benchmark on Snapdragon X Elite
Jun 9, 2026Zhiyuan Cheng, Longying LaiAgentic Retrieval-Augmented Generation Systems Neural Processing Units
School of Engineering Stanford University Stanford, CA, USA · Simon Business School University of Rochester Rochester, NY, USA
CFRNet: Cycle-Consistent Fixed-Point Training for Real-Time Blind Face Restoration on Consumer Embedded NPUs
Jun 5, 2026Fuchen Li, Xinyang Wang, Yahui Zhang +4Blind Face Restoration Image Restoration
University of Florida, Gainesville, FL, USA · University of Southampton, Southampton, UK · Chongqing University, Chongqing, China +2
Uncertainty-Aware End-to-End Co-Design of Neural Network Processors: From Training and Mapping to Fabrication
Jun 3, 2026Yuyang Du, Yujun Huang, Gioele ZardiniCo-Design Neural Processing Units
School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore · Laboratory for Information & Decision Systems, Massachusetts Institute of Technology, Cambridge, MA 02139
When NPUs Are Not Always Faster: A Stage-Level Analysis of Mobile LLM Inference
May 22, 2026Pu Li, Jiawen Qi, Qinyu ChenNeural Processing Units LLM Inference Optimization
Leiden Institute of Advanced Computer Science (LIACS), Leiden University, The Netherlands.
HyperBones: Realtime Bone-driven Neural Garment Simulation with Hypernetwork Conditioning
May 19, 2026Astitva Srivastava, Hsiao-Yu Chen, Ryan Goldade +9Motion-Aware Clothing Dynamics Garments
IIIT Hyderabad, India · Meta Reality Labs, USA · IIT Jodhpur, India
Quant.npu: Enabling Efficient Mobile NPU Inference for on-device LLMs via Fully Static Quantization
May 19, 2026Jinghe Zhang, Daliang Xu, Chenghua Wang +5Post-Training Quantization Neural Processing Units
TurboGR: An Accelerated Training System for Large-Scale Generative Recommendation
May 13, 2026Huichao Chai, Zhixin Wu, Xuemiao Li +8Sequential Recommendation Neural Processing Units
Huawei
Real Image Denoising with Knowledge Distillation for High-Performance Mobile NPUs
May 5, 2026Faraz Kayani, Sarmad Kayani, Asad Ahmed +2Neural Processing Units Mobilenetv2
Computer Vision Lab, CAIDAS & IFI, University of Würzburg, Germany
DPU or GPU for Accelerating Neural Networks Inference -- Why not both? Split CNN Inference
Apr 30, 2026Ali Emre Oztas, Mahir Demir, James Garside +1Neural Network Inference Edge Devices
The University of Manchester
Efficient Mixture-of-Experts LLM Inference with Apple Silicon NPUs
Apr 20, 2026Afsara Benazir, Felix Xiaozhu LinNeural Processing Units Mixture-Of-Expert Inference
University of Virginia Charlottesville, Virginia, USA
Towards Real-Time ECG and EMG Modeling on $μ$ NPUs
Apr 20, 2026Josh Millar, Ashok Samraj Thangarajan, Soumyajit Chatterjee +1Physiological Signals Neural Processing Units
Imperial College London London, United Kingdom · Nokia Bell Labs London, United Kingdom · Brave Software & University of Cambridge London, United Kingdom

Sep 28, 2026, cs.LGJ/K move · Enter open · S save

From Experience to Expertise: Adoption-Aware Memory Learning for Data-Scarce NPU Kernel Synthesis

Longxiao Fan, Tao Zhang, Han Yan, Jiajun Li+4

Fudan University · University of Science and Technology of China · Huawei Technologies Ltd.

Neural Processing Units Cross-Task Knowledge Transfer Credit Assignment

Open paper AskPDF arXiv

High-performance kernels underpin efficient accelerator execution but require expert tuning and lengthy manual optimization cycles. LLM coding agents promise automation, yet their CUDA knowledge transfers poorly to data-scarce domain-specific architectures (DSAs) such as NPUs, whose execution models and memory hierarchies differ substantially from those of GPUs. To address this transfer gap, post-training methods adapt LLMs to NPU programming but depend on scarce expert data and substantial training compute. Memory-learning agents instead adapt through external memory, but their uniform credit assignment gives adopted and unused experiences the same reward target, potentially biasing subsequent retrieval rankings. Moreover, when learned values guide only retrieval, high-value experiences that generalize across operators must be retrieved repeatedly rather than retained in context, thereby increasing retrieval overhead and weakening cross-task guidance. We therefore present SAGE, a persistent self-improving agent for NPU kernel synthesis. Adoption-Traced Utility estimation (ATU) combines explicit adoption records with kernel evaluation outcomes for adoption-aware credit assignment. Utility-Gated Consolidation (UGC) uses positive utility and repeated adoption across operators to select and abstract reusable rules into a bounded resident context. On NPUKernelBench, SAGE achieves a 95.5% execution rate versus 84.1% for the strongest controlled baseline, with 86.9% of solved operators outperforming torch_npu. With GLM-5.3, SAGE achieves a 43.99x speedup over the torch_npu reference on sparse flash attention. These results show that adoption-aware credit assignment and selective consolidation enable agents to accumulate and reuse hardware-specific knowledge across tasks.

Momentum

Related topics

Latest papers 31