Efficient Neural Network Inference

Follow this topic Ask about this topic

Momentum

39 papers in the last four weeks, up 225% on the four weeks before. 0.4% of all new papers.

Jul 13Week of Sep 28

Latest papers 314

Oct 7, 2026cs.CL

SpikingVLA: Asynchronous Spiking Vision-Language-Action Models

ANN-to-SNN conversion offers a practical route toward energy-efficient spiking Vision-Language-Action (VLA) models by bypassing the substantial cost of training large-scale SNNs from scratch. However, existing methods often require many timesteps to maintain competitive performance, resulting in substantial inference latency for real-time VLA deployment. To address this challenge, we introduce SpikingVLA, an ANN-to-SNN conversion framework that enables accurate and low-latency spiking VLA inference. Specifically, we propose a Dendritic Integrate-and-Fire (DIF) neuron that alleviates channel-wise activation outliers through dendritic mixing and adaptive somatic firing, enabling accurate ANN-to-SNN conversion with fewer timesteps. Building on DIF neurons, we further introduce an asynchronous execution mechanism that overlaps temporal computation across VLA components, reducing synchronization overhead and latency. Extensive experiments demonstrate that SpikingVLA achieves competitive navigation performance with substantially improved inference efficiency. Compared with existing spiking VLA methods, SpikingVLA improves SR and SPL by 11.9% and 12.6%, respectively, while reducing first-action latency by 11.2

\times

. These results establish SpikingVLA as a practical framework for deploying pretrained VLA models with high-performance and low-latency spiking inference.

Jingya Wang, Dehao Zhang, Shuai Wang +3

University of Electronic Science and Technology of China · Shenzhen Loop Area Institute · The Chinese University of Hong Kong (Shenzhen)

Efficient VLA Model Inference Spiking Neural Networks Vision-Language-Action Models Efficient Neural Network Inference

Oct 7, 2026cs.CV

EM-SNN: Efficiently Modulated Spiking Neural Network for Remote Sensing Image Dehazing

Although spiking neural networks (SNNs) provide an energy-efficient alternative to artificial neural networks (ANNs), their application to remote sensing image dehazing remains limited. A key challenge arises from the coupling between haze-induced high-frequency attenuation and discrete spike thresholding. This interaction suppresses weak responses and fundamentally limits the recovery of edges, textures, and fine details in spiking dehazing models. To address this challenge, we propose the Efficiently Modulated Spiking Neural Network (EM-SNN), a dedicated spiking framework tailored to remote sensing image dehazing. EM-SNN integrates a statistics-driven Threshold-Modulated Leaky Integrate-and-Fire (TM-LIF) neuron to adaptively compensate for haze-induced contrast compression, together with a Spike Sobel Modulation (SSM) module that enhances structural cues and reduces depth-wise attenuation during spiking feature propagation. By jointly modulating activation scales and structural representations, EM-SNN improves dehazing performance while preserving the inherent event-driven sparsity of SNNs. Experiments on HRSD, RICE, RRSHID, and SateHaze1K demonstrate that EM-SNN achieves competitive dehazing performance while consuming only one quarter of the energy of the strong ANN baseline SFRDP-Net.

Jie Shao, Jiaqi Ma, Wenwen Min +4

Zhongnan University of Economics and Law, Wuhan, China · Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, UAE · Yunnan University, Kunming, China +2

Image Dehazing Spiking Neural Networks Efficient Neural Network Inference

Oct 7, 2026cs.CV

Hardware-aware Calibrated Clustered Attention for Efficient Visual Geometric Transformers

The Visual Geometry Grounded Transformer (VGGT) marks a significant leap forward in 3D scene reconstruction, as it is the first model that directly infers all key 3D attributes (camera poses, depths, and dense geometry) jointly in one pass. However, this joint inference mechanism requires global attention layers with extremely long sequences that causes a significant latency bottleneck. In this paper, we propose blockwise clustered attention (BC attention) to accelerate the global attention layers in VGGT. By limiting the clustering within HW-friendly neighborhood blocks, BC attention reduces the computation overhead of query clustering as well as the costly data movement between on- and off-chip memory. This enables BC attention to scale to long sequences and deliver practical latency improvements on GPUs. Moreover, we introduce a hashing hyperplane calibration method and a threshold-based error compensation method to reduce clustering errors efficiently, which is a bottleneck in the current clustered attention mechanism. Overall, our experiments on GPU demonstrate that calibrated BC attention accelerates the global attention layers by 2.10-2.63

\times

and the whole backbone by 1.77-2.35

\times

with negligible loss (1%) for large scenes. With a small performance loss (< 5%), calibrated BC attention further achieves a 2.26-2.87

\times

latency improvement on the global attention layers and a 1.90-2.55

\times

improvement on the backbone.

Weitian Wang, Shubham Rai, Cecilia De La Parra +1

Robert Bosch GmbH, Germany · Ruhr University Bochum, Germany

Efficient ViTs Attention Mechanisms Efficient Neural Network Inference 3D ViTs 3D Reconstruction

Oct 6, 2026cs.AI

AnyBottle: A Recipe to Only Keep the Concepts You Really Need

Concept bottleneck models (CBMs) make predictions inspectable and intervenable by routing them through human-interpretable concepts, but originally required concept annotations. Annotation-free variants remove this requirement, but typically use large concept vocabularies, static at both training and inference, producing bottlenecks larger than any task or prediction needs and harder to inspect. We propose AnyBottle, a single recipe for building compact, task-specific CBMs. AnyBottle assumes only a frozen backbone and an unsupervised concept pool, such as a sparse autoencoder. A black-box teacher trained on the same backbone then guides selection: each round adds the concept that best explains the bottleneck's current failures, with candidates restricted to regions of teacher/student disagreement. Trained with nested dropout over this selection order, the final bottleneck predicts accurately from any concept prefix, so inference spends fewer concepts on inputs it is confident about early and more on hard ones. Since no stage is modality-specific, a new domain and task requires swapping only the backbone and concept pool. Across six vision and two text datasets and two teacher paradigms, AnyBottle yields bottlenecks with fewer concepts and higher concept consistency than annotation-free baselines, while staying close to the black-box reference. Overall, AnyBottle shows that going annotation-free need not mean going large: a small, discovered vocabulary can be as expressive as a much larger, fixed one.

Wolfgang Stammer, Sukrut Rao, Hevra Petekkaya +2

Max Planck Institute for Informatics, Saarland Informatics Campus · Technical University of Darmstadt

Efficient Neural Network Inference Concept Bottleneck Models

Oct 4, 2026cs.LG

Measuring and Reducing Cross-Vendor Mismatch in Language Models

Running the same language model on different graphics processing unit (GPU) vendors can produce different logits, even when the model weights and inputs are the same. We analyze cross-vendor mismatch in two dense and two mixture-of-experts (MoE) models with five metric families, namely bitwise equality, logit differences, top-K consistency, token agreement, and task accuracy. We trace one source of the mismatch to accumulation order inside vendors' matrix instructions. Upcasting to FP32 reduces the dense model's logit error by 43% at three times the runtime, yet keeping only the MLPs in BF16 retains 94% of this gain at 1.3 times the runtime, so most of the cost of full upcasting buys little. In the MoE models, FP32 and FP16 both lower the probability error but raise the logit error and change expert selection, and FP16 fails in the dense model. An output-head low-rank adapter (LoRA) does not help either, since the final hidden state does not predict the mismatch. The mismatch also carries into training. With every seed fixed, a student distilled from a teacher running on AMD answers 431 MMLU questions differently from one distilled from the same teacher on NVIDIA. Under FP32 upcasting, bitwise equality barely changes while the output distributions move most of the way to the reference, so judging cross-vendor agreement by a single measure misreads both its cost and its gains. Code is available at https://github.com/crova-project/crova.

Erland Hilman Fuadi, Chong Tian, Xiaosong Ma +1

Mohamed bin Zayed University of Artificial Intelligence

Efficient Neural Network Inference ML Reproducibility

Oct 4, 2026eess.IV

TIRMamba: A Thermal-Prior-Modulated State-Space Network for Sub-Million-Parameter Infrared Image Super-Resolution

Infrared image super-resolution is currently led by Mamba-based networks with 26 to 37 million parameters, which are difficult to deploy on the airborne and handheld platforms where thermal imaging is most needed. This paper presents TIRMamba, a network with 896K to 910K parameters for single-channel thermal imagery. A Thermal Prior Highway computes gradient, local-contrast and spectral cues once at the input and, through one adapter per residual group, modulates a weight-tied bidirectional state-space trunk and gates its dual-scale detail branch; a tri-path reconstruction adds the learned residual to a bicubic radiometric baseline. Because the standard benchmark provides only 265 infrared training images and evaluates fusion products on full images, we train with a replay strategy: grayscale DIV2K pre-training followed by fine-tuning on 64-pixel patches drawn with equal probability from the infrared and natural corpora. At scale factor 4, TIRMamba matches the strongest protocol-trained methods on both official test sets with 29 to 40 times fewer parameters and 2.8 to 9.4 times lower latency; at scale factor 2 it gives the highest SSIM on both. A variant with prior-conditioned selectivity, TIRMamba-Rad, corrects a 3 dB raw-thermal failure of an intermediate size-invariant design and gives the best results at scale factor 4 on raw-thermal, unmanned-aerial-vehicle and independent-sensor test sets. Code and trained models will be released at https://github.com/julian135707/TIRMamba upon acceptance.

Chun-An Lin, Tsung-Jung Liu, Yen-Chieh Ouyang

Department of Electrical Engineering, National Chung Hsing University, Taichung 402202, Taiwan · College of Computing and Informatics, Providence University, Taichung 433303, Taiwan

Thermal Imaging Efficient Neural Network Inference Visual SSMs Image Super-Resolution

Oct 4, 2026cs.AR

SparseCraft: Agentic Hardware-Software Co-Optimization for Sparse Computing

Sparse-accelerator design spaces are usually searched against analytical models, so a design point is admitted on what a model predicts rather than on what the hardware does. SparseCraft closes that gap with a language model inside a closed CHIA loop. In each of 15 iterations the model reads the measured outcome of the previous one and edits the Chisel RTL, the memory configuration and the sparse-kernel schedule of a Gemmini accelerator through MCP tool servers, and no candidate counts until it has been checked for legality, elaborated, simulated cycle-accurately, checked bit-for-bit on every output against a golden reference, and synthesised. The harness turns each measurement into the next work order, a diagnosed bottleneck with matching strategy guidance, the history of tried designs and a score of the model's own prediction, and a second model repairs changes that fail a gate. On a

512 \times 512

GraphChallenge sparse-DNN layer the loop reaches 2.1x fewer cycles, 9.8x less off-chip traffic and 22.8% less area than the block-sparse Gemmini baseline, with 5.61x higher modelled perf/W and 11.8x lower EDP. The levers span three layers: a schedule that keeps the dense operand resident removes 9.8x of the traffic, a zero-gated MAC and a zero-row skip unit that the model wrote in Chisel cut energy, and resizing the memories cuts area.

Rajatabha Chakraborty, M P Samartha, Vedant Pahariya +1

International Institute of Information Technology, Hyderabad, India

Efficient Neural Network Inference Large Language Model-Guided Optimization Hardware-Software Co-Design

Oct 1, 2026cs.LG

Distillation of Tabular Foundation Models into Efficient Predictors

Tabular foundation models (TFMs) achieve strong predictive performance through in-context learning, yet repeatedly conditioning on labeled data makes inference expensive. Knowledge distillation can reduce this cost by transferring their predictive ability to lightweight, dataset-specific students. However, the dependence of TFM predictions on both a labeled context and a query introduces two design questions: how to construct teacher supervision and whether expanding query coverage improves distillation. We examine these questions across two TFMs and both neural and tree-based students, and derive an effective distillation recipe. The recipe uses the full labeled training set as teacher context and trains students solely on teacher predictions for observed and synthetic queries. On TabArena, the resulting students outperform their supervised trained tuned-and-ensembled counterparts by 57-98 Elo points. Applied unchanged to TALENT, the same recipe improves matched default students on 236-258 of 300 datasets and reduces median primary error by 4.0-6.4%. The distilled students also achieve median inference speedups of 3.0-21.6 times over their teachers, offering a practical trade-off between predictive performance and repeated inference cost. Code is available at https://github.com/nums-ai/TFM_Distillation .

Minho Jeong, Dooho Lee, Jinmo Lee +1

Nums AI

Tabular Foundation Models Efficient Neural Network Inference Tabular ICL Knowledge Distillation

Sep 30, 2026cs.LG

Random Recursive Models

Recursive models create computational depth through parameter reuse, offering a parameter-efficient alternative to increasing model size. However, most recursive models repeatedly apply one learned transformation or a prescribed sequence of transformations, restricting computation to a fixed layer order. We introduce the Random Recursive Model (RRM), which maintains a pool of

L

learned layers and performs

T

recursive steps by sampling one layer independently with replacement for each example and step. This enables flexible layer reuse while retaining the parameter efficiency of recurrence. We evaluate RRM on challenging reasoning tasks, where it matches or exceeds the baselines, often with 50-75 % fewer parameters. RRM can vary its depth at inference, including beyond that seen during training, without retraining or adding parameters, improving tasks that benefit from deeper iterative computation. RRM also supports Monte Carlo inference and probabilistic test-time scaling, both of which improve performance without retraining. These insights may open new directions in neural network architecture design.

Jama Hussein Mohamud, Mirco Ravanelli

Mila – Quebec AI Institute · Université de Montréal · Concordia University

Efficient Neural Network Inference LLM Reasoning Recursive Language Models

Sep 30, 2026cs.LG

A Tilted Bowl Is Not a Slippery Slope: Compressing Looped Models

Looped models reason by applying the same block of weights many times, so compressing that block saves memory traffic on every loop. Compressed looped models, however, often collapse, and the collapse is usually blamed on rounding error that accumulates from loop to loop. In this work we test that account on more than 30 models from five families and find, to our surprise, that it holds only for loops that never settle. When a loop settles, a fixed rounding error does not accumulate. It moves the point where the loop settles, much as tilting a bowl moves where a ball comes to rest, and the answer is lost only when the shift is larger than the readout tolerates. This picture lets us predict which models fail from a single label-free measurement, and it tells us why failed models recover: their loops still settle, so a few final loops with 8-bit weights bring the answer back. Motivated by these findings, we build a controller that stops when the model's halting head fires and then finishes with 8-bit loops. On Sudoku-Extreme and Maze-Hard it beats fixed-depth inference by up to 15 points under a third of the weight traffic.

Steven Kolawole, Pearse Jim, Opegbemi M. Busoye +2

Carnegie Mellon University · ML Collective

Efficient Neural Network Inference Recurrent Neural Networks Neural Network Quantization Neural Network Compression Adaptive Computation

Sep 29, 2026cs.DC

RLX: A Unified Multi-Backend Tensor Compiler and Distributed Runtime in Rust

Production machine learning (ML) stacks often split graph compilation and kernel execution across different layers and languages, making backend behavior, deployment guarantees, and performance fallbacks hard to reason about end-to-end. RLX addresses this gap with a single Rust codebase that combines compiler and runtime roles around one primitive-level, three-level intermediate representation (IR), plus a transparent dispatch contract that resolves each operator to native, common-IR, or rewritten lowering and fails compilation when legalization is not possible. The same IR targets fourteen runtime devices (cpu, metal, mlx, ane, cuda, rocm, oneapi, tpu, hexagon, gpu, vulkan, opengl, directx, webgpu) and two specialty codegen paths (Cortex-M INT8 and FPGA), ingests safetensors, GGUF, ONNX, and rten formats, supports F16/BF16/F64/C64 and quantized INT4/INT8 flows with AMP/PTQ/QAT, and scales via tensor-/pipeline-parallel collectives over TCP and RDMA transports. Beyond neural workloads, RLX also extends to scientific/physics-style domains through sparse and dense linear algebra extensions (e.g., CSR LU/CG/matvec and LAPACK- backed factorizations) and 3D Gaussian splatting operators. We evaluate RLX against PyTorch, TensorFlow, JAX, candle, burn, tch, rten, MLX, CoreML, IREE, Glow, TensorRT, and tinygrad under identical input generation and p50 measurement methodology on one host. On all-MiniLM-L6-v2, RLX-Metal is fastest at every batch (e.g., 16.6 ms at batch 32 vs. PyTorch-MPS 26.7 ms). In the MNIST training table, RLX also has the top-throughput entry (graph-fused MLP: 946,487 img/s), above NumPy+BLAS (787,349 img/s), while retaining 100% top-1 parity on reference checks (e.g., Qwen3).

Eugene Hauptmann, Nataliya Kosmyna

Massachusetts Institute of Technology Cambridge, MA, USA

Efficient Neural Network Inference

Sep 28, 2026cs.CV

Hardware-Aware Functional Kolmogorov-Arnold Networks for Efficient Medical Image Enhancement and Segmentation

Functional Kolmogorov-Arnold Networks (FunKAN) achieve state-of-the-art accuracy on MRI Gibbs artifact removal and anatomical segmentation, but their 11.6 M parameters and 8.7 GFLOPs are too large for edge medical devices. We present FunKANLite, a two-stage, hardware-aware compression of FunKAN for point-of-care use. FunKANLite-TR reduces the spatial prior and replaces the ResBlock offset predictor with a depthwise-separable block. It has 1.9x fewer parameters than FunKAN and no loss in accuracy. We then distill FunKANLite-TR into FunKANLite-ST, which lowers the Hermite basis rank, factorizes the spatial prior into a low-rank form, and halves the filter widths. FunKANLite-ST has 5.6x fewer parameters and 3.7x fewer GFLOPs than FunKAN. It stays within 1.4 percentage points IoU of FunKAN on BUSI, GlaS, and CVC-ClinicDB, and reaches 33.95 dB PSNR on IXI. On an NVIDIA Jetson Orin Nano and a Raspberry Pi 5, FunKANLite-ST reduces energy per inference by up to 68% and raises throughput by 2.9x.

Mohammad Sadegh Sirjani

Department of Computer Science University of Texas at San Antonio, San Antonio, TX, USA

Efficient Neural Network Inference Kolmogorov-Arnold Networks Image Restoration Medical Image Segmentation Neural Network Compression

Sep 28, 2026cs.LG

Fiona: Accelerating FHE Inference with Packing-Aware Ternary Weights

Fully homomorphic encryption (FHE) enables neural network inference directly on encrypted inputs, but it remains orders of magnitude slower than plaintext in- ference. Applying the server's plaintext weights to encrypted activations involves plaintext-ciphertext multiplications (PMult) and accounts for more than half of inference time in recent systems. Ternary quantization can replace these multipli- cations with additions and subtractions, but the savings rarely materialize under packed execution. A single PMult applies a weight group fixed by the packing layout and can be avoided only when all its weights share the same ternary value. Ternarizing all groups, however, largely degrades accuracy. We present FIONA, an offline optimizer that selectively ternarizes weights within a given packing layout based on the estimated effect of ternary conversion on the model's performance. FIONA encourages a shared ternary value within each weight group and retains full-precision weights for sensitive groups, so ternar- ized and full-precision paths coexist within a layer. It then compiles these hybrid operators exactly, applying common scaling factors once to accumulated inputs and reusing sums across outputs. Weight ternarization can also narrow the input ranges of downstream polynomials. FIONA fits lower-degree replacements under a cumulative accuracy budget, reducing multiplicative depth and bootstrapping. On VGG11, ViT, and BERT, FIONA reduces PMult operations by 53.4-79.5% and accelerates end-to-end encrypted inference by 2.38x, 1.68x, and 1.84x, re- spectively, with less than 1% accuracy loss across all three models.

Yiteng Peng, Zhibo Liu, Dongwei Xiao +1

The Hong Kong University of Science and Technology · State Key Laboratory of Novel Software Technology, Nanjing University

Mixed-Precision Quantization Efficient Neural Network Inference Ternary Quantization Homomorphic Encryption Weight-Only Quantization

Sep 28, 2026cs.LG

Latency and accuracy tradeoffs in Spiking Neural Networks

Spiking neural networks are attractive for low-power speech command recognition, yet their latency has received far less attention than their energy efficiency, and their multi-timestep execution is widely assumed to make them slower than quantized neural networks. This paper challenges the assumption that more local timesteps necessarily imply higher network latency. By overlapping computation across adjacent layers at the timestep level, SNNs may complete execution in less time than comparable bit-serial QNNs. However, this overlap relies on spikes firing on incomplete inputs, and a spike once generated cannot be withdrawn, so its error persists and reduces accuracy. Waiting for more input before firing would seem to improve accuracy at the cost of reduced overlap. Yet we find and prove that this intuition fails at some layers, where even a small increase in waiting can change spike timing and downstream computation, making the network both slower and less accurate. We therefore propose a Pipeline Delay Search method which selects each layer's delay by balancing task-level accuracy gains against added network latency. We then adapt the selected configurations through spike-based quantization-aware training and bounded tuning of firing thresholds and initial membrane potentials. Together, these steps form Falcon, a framework for Fine-grained Analysis of Latency and Controlled firing which systematically analyzes and optimizes SNN latency under a spatial analog compute-in-memory mapping with shared digital engines. We evaluate Falcon on GSCV2 and SSC, achieving competitive accuracies of 96.31 and 83.02 at modeled network-core latencies of 119.64 and 124.00us, respectively. Together, our analysis and results show that SNNs can compute more yet finish faster, and wait longer yet predict worse, highlighting why Falcon matters for both latency and accuracy.

Zhanglu Yan, Zixuan Zhu, Kaiwen Tang +3

Department of computer science, National University of Singapore · Shanghai Advanced Research Institute, Chinese Academy of Science · University of Chinese Academy of Sciences, Beijing +1

Efficient Neural Network Inference Automatic Speech Recognition Spiking Neural Networks

Sep 28, 2026cs.CV

From UNI2-h to ConvNeXt-T: Lightweight Nuclei Instance Segmentation via Knowledge Distillation

Nuclei instance segmentation is a core task in digital pathology, yet high-accuracy models rely on large vision transformer (ViT) encoders whose inference speed cannot meet real-time clinical demands. We propose a lightweight scheme that distills the UNI2-h pathology foundation model into a ConvNeXt-Tiny student (Ours-T, 34.7M parameters, 1/20 of the teacher) via output-level knowledge distillation. Ours-T achieves an mPQ of 0.519 on PanNuke (98.8% of the teacher), a zero-shot bPQ of 0.668 on MoNuSeg, and an inference speed of 634.3 img/s, requiring only 0.045 s for full-resolution 1024^2 analysis (21.8x speedup). Experiments further show that multi-scale gated convolution (MALA) yields no gain under ViT encoders, and output-level distillation alone suffices for efficient knowledge transfer.

Wenyan Li

School of Computer Science, Wuhan University, Wuhan, China

Efficient Neural Network Inference Histopathology Image Segmentation Instance Segmentation Cross-Architecture Knowledge Distillation Knowledge Distillation

Sep 28, 2026cs.LG

SpikeLite: Lightweight Spiking Neural Networks for Time-Series Forecasting

Spiking neural networks (SNNs) offer an energy-efficient paradigm for time-series forecasting through spike-driven computation. However, recent SNN forecasters often pursue higher accuracy through increasingly complex attention mechanisms, or specialized neuronal dynamics, weakening the lightweight motivation of SNNs. We introduce SpikeLite, a spiking forecasting framework built around two modules: a Frequency-Selective Spiking Encoder (FSSE) for frequency-sensitive temporal encoding and a Sparse Spiking Channel Attention (SSCA) module for selective cross-channel interaction. FSSE exploits the low-pass filtering behavior of LIF dynamics to reorganize each input sequence into frequency-sensitive components while collectively preserving the input at the decomposition stage. SSCA then learns a binary mask from encoded channel representations and uses it to selectively exchange information within spike-driven self-attention, retaining informative cross-channel interactions while suppressing redundant ones. When explicit channel interaction is unnecessary, SpikeLite uses the lighter FSSE-only channel-independent path. Experiments under the SeqSNN and SpikF protocols cover four standard multivariate and eight long-term forecasting benchmarks. SpikeLite achieves the best aggregate performance under both protocols, with an average

R^2

of 0.790 and RSE of 0.440, and lowest average MSE/MAE of 0.343/0.345 in long-term forecasting. Moreover, evaluation on the ECL dataset shows that SpikeLite achieves the lowest reported energy consumption, further demonstrating its potential for energy-efficient time-series forecasting.

Bang Hu, Changze Lv, Mingjie Li +3

School of Computer Science, Fudan University, Shanghai, China

Multivariate Time Series Forecasting Efficient Neural Network Inference Time Series Forecasting Spiking Neural Networks Sparse Attention

Sep 28, 2026cs.LG

Reference-Tail Trust:Certified Probability Floors for Learned Updates Inside a Deployed Network

Graph neural networks (GNNs) need to exploit improved message passing without surrendering control over predictions already trusted in deployment. We introduce Reference-Tail Trust (RTT), a framework that admits learned updates inside a frozen GNN and certifies the prediction actually served. RTT couples graph-based proposal states with a constrained internal optimizer: each displacement is charged for its worst-case terminal cross-entropy increase through the incumbent's remaining message-passing layers. A trajectory-validated tube and an independent checker enforce per-node probability floors,

p^{\mathrm{s}}_{ic} \ge e^{-H_{\mathrm{row}}} p^{\mathrm{r}}_{ic}

, and a call-level budget,

\sum_i w_i D_\infty(p^{\mathrm{r}}_i \| p^{\mathrm{s}}_i) \le H^+

, uniformly over labels. Calls whose adapted outputs pass certification require no separate full incumbent rollout; failed certificates trigger whole-call fallback. We derive the exact probability-floor frontier by water-filling, characterize architecture-constrained efficiency, and establish conditions under which internal propagation exploits evidence unavailable to restricted output correctors. In the reported ogbn-arxiv audit, RTT achieves

6.5\times 10^{-3}

nats of mean gain per call, with a one-sided 95% regression-rate upper bound of 0.95% and a 95% negative-flip upper bound of 0.51% on the uninspected part of the reserved node population. Its mean gain is 61% of a cross-fitted posterior-based frontier estimate and exceeds the strongest matched one-pass corrector by

+0.9\times 10^{-3}

nats. Reported experiments span eight proposals, six graph-incumbent families, structural and temporal graph shifts, and molecular prediction, with additional image and tabular evaluations. RTT makes GNN adaptation a budgeted, certifiable inference decision rather than an unconditional model replacement.

Abdolvahab Khalili Sadaghiani, Jose Nunez-Yanez

Efficient Neural Network Inference Graph Neural Networks

Sep 28, 2026cs.LG

FlexLoop: Depth-Elastic Looped Policies for Adaptive Test-Time Computation in Deep RL

Looped architectures scale computation by reusing the same parameters across recurrent steps, and recent work shows that they substantially improve deep reinforcement learning policies on long-horizon tasks. Since recurrent depth directly controls computation, one may expect looped policies to naturally support elastic inference across recurrent depths. Surprisingly, we find that pretrained looped policies exhibit severe recurrent-depth specialization: reliable decisions are concentrated near the full trained depth, tying deployment computation to this depth even when less computation may suffice. Achieving depth elasticity, i.e., reliable decisions across recurrent depths with adaptive computation at deployment, therefore remains a key challenge. To address this, we propose FlexLoop, a novel post-training framework that converts pretrained fixed-depth looped policies into depth-elastic policies. FlexLoop keeps training on the original RL objective to preserve full-depth capability while performing adjacent-depth policy distillation to progressively transfer decision quality from deeper to shallower recurrent steps. The resulting policy supports reliable inference across recurrent depths and enables state-wise adaptive inference through recurrent-depth consistency. Experiments on

30

online and offline long-horizon goal-conditioned environments show that FlexLoop preserves full-depth performance while making shallower depths effective. Keeping competitive performance, FlexLoop reduces average recurrent depth by up to

\bf{43\%}

and achieves up to

\bf{1.34\times}

wall-clock speedup in a stress test.

Xun Wang, Ruishuo Chen, Yu Chen +2

Institute for Interdisciplinary Information Sciences, Tsinghua University

Efficient Neural Network Inference Reinforcement Learning Recurrent Neural Networks RL Post-Training

Sep 28, 2026cs.LG

EntroPack: Fast and Accurate Entropy-Coded Weight Compression at Arbitrary Bitrates

Weight compression helps large neural networks fit deployment memory budgets, but common fixed-width formats offer only coarse storage choices. Entropy coding supports finer rates, yet the achieved size depends on the quantized weight distribution and coding overhead. Exploiting this flexibility requires accurate rate selection and efficient weight reconstruction for inference. We present EntroPack, an entropy-coded weight compressor that supports arbitrary target bitrates without activation calibration or fine-tuning. It combines row-normalized

E_8

lattice quantization with a conditional probability model of lattice coordinates. Sampled storage estimates select the quantization resolution without repeated full-stream encoding. The final coordinates are entropy-coded in independently decodable tiles, enabling fast, fused symbol decoding and numerical weight reconstruction on the GPU. EntroPack supports floating-point and integer weight containers, such as BF16, FP16, FP8, and INT8, with storage bitrate controlled independently of numerical precision. Online decoding adds latency that grows with weight count, making the method well suited to compute-intensive workloads such as diffusion denoising and Transformer prefill. Experiments demonstrate fast encoding and modest inference overhead in these settings. When compressing the linear-layer weights of the image generator Z-Image-Turbo, EntroPack achieves substantially lower weight and denoiser output errors than fixed-width formats at comparable storage rates, with modest denoising-step overhead. Targeting 4 bits per parameter, it achieves lower weight and denoiser output errors than NF4, including about 24% lower relative

L_2

weight error, with less storage. Source code is available at https://github.com/modelscope/entropack.

Hong Zhang, Zhongjie Duan, Yingda Chen

Alibaba Inc.

Efficient Neural Network Inference Weight-Only Quantization Neural Network Compression Entropy Coding

Sep 28, 2026cs.CV

Analytical and Convolutional Neural Network-Based Motion-Vector Propagation for Efficient Video Object Detection

Continuous video analytics requires accurate localization at low latency within embedded power budgets. This paper presents a hardware-software design methodology that reuses codec motion vectors (MVs) between detector invocations. Two alternative models support translation and scale changes: analytical motion-vector propagation (Analytical-MV) and learned propagation using a convolutional neural network (CNN) (CNN-MV). The learned model uses convolutional operations and independent object updates suited to parallel execution on an edge graphics processing unit (GPU). Analytical-MV combines a harmonic-mean precision-recall score (F1) of 0.909 with a mean end-to-end latency of 9.03 ms and an energy consumption of 0.177 J per frame, yielding the lowest latency and energy among the evaluated configurations. Relative to detection on every frame, it reduces mean latency by 25.9% and energy per frame by 36.4%. CNN-MV offers a different trade-off: its fastest configuration raises recall from 0.871 for Analytical-MV to 0.890 and lowers mean power from 19.64 to 17.32 W, while achieving a latency of 18.42 ms and an energy consumption of 0.319 J per frame. It is therefore useful when recall or operating power is more important than minimum latency and energy. Execution on a deep learning accelerator (DLA) further reduces time-averaged GPU utilization relative to GPU execution. Host-processing optimization substantially improves both latency and energy, demonstrating the value of jointly designing temporal models and their execution pipelines.

Ashiyana Abdul Majeed, Mahmoud Meribout, Neethu Joseph

Department of Computer and Information Engineering, Khalifa University, Abu Dhabi, UAE

Efficient Neural Network Inference Edge Computing Video Object Detection Convolutional Neural Networks

Sep 27, 2026cs.LG

Task-Aware Discretization of Differentiable Logic Gate Networks

Differentiable logic gate networks (DLGNs) enable gradient-based training of highly efficient Boolean networks by relaxing discrete logic gates during training and discretizing them for inference. Standard approaches make this discretization decision locally, typically through argmax selection and confidence- or entropy-based convergence criteria. We show that local discretization can be task-suboptimal even for globally optimal relaxed solutions, with high gate confidence providing no general guarantee, and derive bounds relating task-aware gate selection to tractable interventions in the relaxed network. Motivated by these results, we study first-order downstream task information for progressive discretization and characterize when this local approximation is reliable. Experiments on convolutional DLGNs reveal a strong locality dependence: first-order scores become unreliable when directly optimized over nonlocal interventions, but accurately assess local argmax decisions for progressive freezing.

Thore Gerlach

Independent Researcher

Efficient Neural Network Inference

Sep 24, 2026eess.AS

Does per-frame early exit pay? A compute-matched study of dynamic depth for on-device speech enhancement

Deep learning-based speech enhancement is increasingly deployed on-device in hearing aids, headsets, and earbuds. Most of these devices, however, can only accelerate static int8 graphs, so a depth-varying network must be implemented as several graphs, orchestrated by a policy. In this paper, we supervise every intermediate depth of one causal model, then we fine-tune its output heads to guarantee that deeper outputs are never worse than shallower ones. Using this training protocol, we can derive a family of static models that are more Pareto-efficient than their equivalently-sized counterparts trained from scratch on the same budget. Specifically, we achieve up to 0.11 higher PESQ for equivalent compute, and match the best PESQ at 30% less compute. We then quantize the models to int8 and measure the latency-quality frontier on an STM32N6 microcontroller. On VoiceBank-DEMAND, the dynamic enhancer lies on the same frontier as the static models, rather than trading quality for dynamic execution. Running the policy on the companion Cortex-M55 takes only 26

μ

s per frame, while splitting the enhancer into separate NPU graphs adds 2.2% latency overhead. The cost of dynamic execution is therefore small.

Clément Laroche, Riccardo Miccini

GN A/S, Denmark · Technical University of Denmark, Denmark

Efficient Neural Network Inference Speech Enhancement Early-Exit Neural Networks On-Device Inference

Sep 24, 2026eess.AS

Beyond Model Size: Redesigning LiSenNet for embedded speech enhancement

Deploying real-time speech enhancement on resource-constrained devices requires meeting strict latency, memory, and energy constraints. Microcontroller NPUs can accelerate neural inference under these constraints, but only through a restricted set of operators in static, integer-quantized graphs. Recent speech-enhancement networks have reduced parameter counts and MACs to levels nominally suitable for microcontrollers, but their operators and execution patterns often remain incompatible with restricted NPUs. We address this gap by redesigning LiSenNet, a 37k parameter sub-band dual-path model, for the STM32N6570-DK Neural-ART accelerator. We replace its recurrent bottleneck with convolutional frequency and temporal mixers, reformulate unsupported operations as static int8-compatible primitives, and use bounded decoder activations to preserve quality after quantization. On VoiceBank-DEMAND, the final NPU-compatible model matches or exceeds the recurrent LiSenNet baseline, reaching PESQ 3.08 versus 3.01 in FP32 and 3.01 versus 2.93 in int8. Deployed on a microcontroller, it processes each 16 ms input hop in 4.83 ms, corresponding to a real-time factor of 0.30. Stateless receptive-field recomputation is an order of magnitude slower at the same frame rate despite higher accelerator utilization. These results show that parameter count and operator compatibility, quantization range, and persistent streaming state must be co-designed to achieve efficient real-time speech enhancement on restricted NPUs.

Clément Laroche, Rasmus Kongsgaard Olsson

GN A/S, Audio Research, Lauptrupbjerg 7, 2750, Denmark

Efficient Neural Network Inference Speech Enhancement Neural Network Quantization

Sep 22, 2026cs.CV

GTR: Gated Token Recurrence for Efficient Dense Prediction

Self-attention-based vision backbones perform well on dense prediction, but the quadratic computational cost of global softmax attention limits their efficiency as image resolution increases. We introduce Gated Token Recurrence (GTR), a softmax-free recurrent vision backbone that combines gated linear attention, alternating spatial scan directions, and spatially enhanced SwiGLU blocks. GTR is distilled from a detection-specialized DINOv3 teacher using only final-layer patch-token alignment through a linear projection and squared

\ell_2

loss, without masked-token prediction or intermediate-layer supervision. With Objects365 detector pre-training, GTR-L achieves 58.9 box AP on COCO \texttt{val2017} with 1.908,ms median batch-one latency under compiled FP16 execution on an RTX4090. The same backbone also transfers to instance segmentation, pose estimation, oriented detection, semantic segmentation, and monocular depth estimation. In an isolated kernel benchmark, our specialized chunkwise CUDA operator is $4.0\times$ faster than FLA v0.5.0 at 1.6K tokens on RTX4090. TensorRT deployment on DRIVE AGX Thor achieves 2.282--8.769,ms median batch-one latency across the evaluated models. These results show that recurrent token mixing can provide an efficient alternative to global softmax attention for high-resolution dense prediction and edge deployment. Project page: https://intellindust-ai-lab.github.io/projects/GTR/

Zhe Feng, Longfei Liu, Wei Liu +7

Didi International Business Group · Intellindust AI Lab · Institute of Automation, Chinese Academy of Sciences +2

Efficient Neural Network Inference Gated Attention Edge Inference Recurrent Transformers Efficient ViTs Linear Attention+1

Sep 21, 2026cs.SD

Narrowband Voice Communication Using Streaming Neural Compression

Low-bitrate speech communication on resource-constrained edge devices remains challenging due to stringent computational, memory, and bandwidth constraints. We present TinyCall, a lightweight neural audio codec designed for real-time speech communication on low-power platforms such as the ESP32 microcontroller and Raspberry Pi. The proposed system targets emergency communication and other bandwidth-limited scenarios while preserving speech intelligibility, speaker identity, and vocal expressiveness. To enable efficient deployment, we propose a minimal neural audio codec architecture together with a framework for converting a causally trained codec into a truly streamable codec through pseudo-lookahead decoding and decoder-input caching. We further replace conventional residual vector quantization (RVQ) with Residual Finite Scalar Quantization (RFSQ) to reduce inference complexity on edge processors and employ a progressive three-stage training strategy for stable optimization under latent quantization. An MFCC-based perceptual loss encourages preservation of speaker characteristics, including harmonic structure and vocal timbre. Experimental results demonstrate real-time operation on a Raspberry Pi 3 while achieving intelligible speech reconstruction at bitrates as low as 2.3 kbps. The proposed approach demonstrates that practical neural speech communication is feasible on highly resource-constrained edge devices.

Dahong Luo, Anannya Trehan, Aritrik Ghosh +1

University of Maryland, College Park

Efficient Neural Network Inference Neural Audio Codecs Speech Processing

Sep 21, 2026cs.LG

Q-DEQ: Discrete Solving and Quantization for Deep Equilibrium Models in Time Series Forecasting under Edge Deployment Coding Constraints

Edge deployment motivates forecasting models with compact parameter storage and low-bit representations. Deep equilibrium models (DEQs) obtain implicit depth by repeatedly applying a shared layer, reducing the parameter cost of explicit layer stacking. Their usual Anderson solver, however, searches for update coefficients in the continuous real domain. We propose Q-DEQ, which formulates local updates in DEQ forward solving as discrete optimization problems. Candidate directions are constructed from the current state and iteration history, and a local quadratic residual model is used to evaluate their combinations. Binary encoding of the direction coefficients yields a quadratic unconstrained binary optimization (QUBO) problem that can be solved by simulated annealing (SA) or a coherent Ising machine (CIM). After fixed-point solving, a re-forward pass applies W8A8 fake quantization to the shared layer's weights and activations. We evaluate Q-DEQ with an iTransformer backbone on five multivariate time series forecasting datasets. Relative MSE differences from the explicit multi-layer baseline range from

-1.16\%

to

+2.90\%

, with lower MSE on two datasets. DEQ parameter sharing reduces parameter counts by factors of

1.80\times

--

3.82\times

; combined with W8A8, static weight storage is reduced by factors of

4.3\times

--

12.8\times

. Local QUBO problems solved using CPU-based SA and the Kaiwu CIM physical backend produce closely matching downstream forecasts. These results establish local discrete solving as a viable component of DEQ time series forecasting and provide a route for executing fixed-point updates through different combinatorial optimization backends.

Ruotong Yang, Hongdong Zhu, Qi Gao +3

Multivariate Time Series Forecasting Efficient Neural Network Inference

Sep 16, 2026cs.LG

Radio-Frequency Convolutional Neural Networks

Running artificial intelligence (AI) models directly on edge devices such as smartphones, wearables, and drones offers low latency, pervasive scalability, and data privacy, but these devices rarely carry the computing capability that modern neural networks demand. Edge accelerators have been developed in response, yet each adds computing hardware to devices already constrained in size, weight, power, and cost (SWaP-C). An alternative lies in what these devices already carry: the frequency mixer in every wireless radio multiplies signals in time, natively performing convolution in the frequency domain. Here we introduce radio-frequency convolutional neural networks (RF-CNNs), which repurpose existing communication hardware for CNN inference. Multi-channel convolutions are mapped onto frequency tones for a passive mixer to execute in a single pass. We experimentally demonstrate that RF-CNN runs deep CNNs up to 26.4 million parameters and nine layers from classification of wireless signals and images to controllable image generation, close to full-precision performance. Because the weights arrive over the air and the analog hardware is shared with communication, the edge device spends energy only on data preparation and readout-down to 0.72 femtojoules per multiply-accumulate, two orders of magnitude less than it would cost on an added digital processor. These results suggest that deployed wireless infrastructure can bring efficient, state-of-the-art AI inference to the billions of devices it already connects.

Zhihui Gao, Shi-Yuan Ma, Yiran Chen +2

Department of Electrical and Computer Engineering, Duke University, Durham, NC 27708, USA. · Research Laboratory of Electronics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA.

Efficient Neural Network Inference Edge Inference Convolutional Neural Networks

Sep 16, 2026cs.CV

Towards Transparent Diagnostics: Investigating Architectural Trade-offs and Explainability in Malaria Detection

More than 80 countries have reported malaria cases with 610 thousand deaths and are projected to increase. Identifying malaria early and accurately helps save lives and effective way to diagnose malaria is through microscopic methods that are labor intensive and require experts with special equipment. Deep learning (DL) has shown promising results in medical diagnosis. Here, we explored various DL models: ResNet18, MobileNetV2, EfficientNet-B2, VGG19 and proposed model ResNet18+TTA (ResNet18 backbone with modified classification head and test time augmentation) for detecting malaria presence using blood smears taken from the NIH Malaria dataset. Our experiment shows MobileNetV2 achieved 96.85 % accuracy with smallest model size (8.49 MB) and fastest inference (1.35 ms). The ResNet18+TTA model achieved 97.96 % accuracy, 0.996 AUC with longest inference time (13.32 ms). Larger architecture outputs a larger model size with moderate accuracy. Upon further pruning, ResNet18+TTA model gained a slight improvement in accuracy and reduced inference time. GRAD-CAM, SHAP and LIME provide explainable AI (XAI) insights into model predictions, using explanation agreement and divergence to evaluate predictive reliability.

Suman Kunwar, Avishek Dangol

DWaste, Baltimore, USA

Efficient Neural Network Inference Medical Diagnosis Explainable Medical Image Analysis

Sep 16, 2026astro-ph.SR

Physics-Informed Neural Networks for Fast Multilayer Spectral Inversion of Hα 6562.8 A and Ca II 8542.1 A Spectra

Strong chromospheric absorption lines such as H

α

6562.8 A and Ca II 8542.1 A provide vital diagnostics of plasma dynamics and thermal structure in the solar chromosphere. Multilayer spectral inversion (MLSI) offers a physically interpretable framework for modeling these lines using a finite number of radiative-transfer layers, but conventional MLSI relies on pixel-by-pixel nonlinear least-squares fitting, making it computationally expensive for large imaging spectroscopic data sets. Here, we introduce a physics-informed neural-network (PINN) framework to accelerate MLSI while preserving its analytic radiative-transfer formulation. The network predicts MLSI parameters directly from observed line profiles and passes them through a differentiable MLSI forward model to synthesize spectra. Training follows a two-stage approach: an initial stage optimized solely via spectral reconstruction loss, followed by fine-tuning that combines spectral consistency with parameter-space supervision from conventional MLSI results on a single reference image. This strategy eliminates the need for large precomputed training sets while maintaining physical interpretability. Applied to Fast Imaging Solar Spectrograph (FISS) observations from the Goode Solar Telescope (GST) targeting both quiet-Sun and active-region regions, MLSI-PINN parameter maps reproduce the primary spatial structures of direct inversions, achieving an arithmetic mean pixel-wise Pearson correlation coefficient of 0.933 across all evaluated parameters. The reconstructed spectra closely match both observed profiles and conventional MLSI fits. Post-training, MLSI-PINN processes a raster in approximately 5-15 seconds compared to 3-5 minutes for conventional MLSI, delivering an inference speedup of about 12-60 times without substantial loss in reconstruction quality, enabling efficient MLSI analysis on large chromospheric data sets.

Ziyang Zhang, Qin Li, Vasyl B. Yurchyshyn +4

Department of Mechanical and Industrial Engineering, New Jersey Institute of Technology, Newark, NJ, USA · Department of Physics, New Jersey Institute of Technology, Newark, NJ, USA · Big Bear Solar Observatory, New Jersey Institute of Technology, Big Bear City, CA, USA +1

Efficient Neural Network Inference Physics-Informed ML Inverse Problems

Sep 15, 2026cs.AR

OptiPrime: Optimizing Private Inference through Protocol-Hardware Co-design

Private deep neural network (DNN) inference based on hybrid homomorphic encryption (HE) and multi-party computation (MPC) can protect user data with a formal guarantee, but at the cost of significant latency overhead due to HE. Customized HE accelerators have been proposed and have achieved orders-of-magnitude speedup for individual HE operations. However, when directly applying a commercial HE accelerator to state-of-the-art HE-MPC frameworks, we observe only limited end-to-end performance gain. This is because HE-MPC frameworks often require wireless transmission of input and output ciphertexts for each HE operation, leading to a severe network communication bottleneck. To overcome this challenge, we introduce OptiPrime, a protocol-hardware co-optimization framework for efficient private DNN inference. OptiPrime features a novel HE protocol for convolutions that substantially reduces the number of transmitted output ciphertexts and mitigates the network communication bottleneck. Meanwhile, as the new protocol introduces complex computation for fewer output ciphertext, we observe new memory access challenges due to a high volume of weight plaintexts and intermediate ciphertexts. Hence, we further propose a lightweight compression system for the weight plaintexts, reducing memory traffic by 10 times, as well as a specialized dataflow to maximize on-chip data reuse of intermediate ciphertexts. Extensive experiments show that our framework outperforms the Cheetah baseline by at most 5.7 times on CPUs and 4.2 times with an accelerator.

Jiangrui Yu, Ye Yu, Si Chen +5

Peking University, Beijing, China · Open Security Research, Shenzhen, China · Tsinghua University, Beijing, China

Efficient Neural Network Inference Secure Multi-Party Computation AI Accelerator Inference Homomorphic Encryption Privacy-Preserving ML