Spiking neural networks (SNNs), inspired by biological neural mechanisms, represent a promising neuromorphic computing paradigm that offers energy-efficient alternatives to traditional artificial neural networks (ANNs). Despite proven effectiveness, SNN architectures have struggled to achieve competitive performance on large-scale speech processing tasks. Two key challenges hinder progress: 1) the high computational overhead during training caused by multitimestep spike firing and 2) the absence of large-scale SNN architectures tailored to speech processing tasks. To overcome the issues, we introduce the input-aware multilevel spikeformer (IML-Spikeformer), a spiking transformer architecture specifically designed for large-scale speech processing. Central to our design is the input-aware multilevel spike (IMLS) mechanism, which simulates multitimestep spike firing within a single timestep using an adaptive, input-aware thresholding scheme. IML-Spikeformer further integrates a reparameterized spiking self-attention (RepSSA) module with a hierarchical decay mask (HDM), forming the HD-RepSSA module. This module enhances the precision of attention maps and enables modeling of multiscale temporal dependencies in speech signals. Experiments demonstrate that IML-Spikeformer achieves word error rates (WERs) of 6.0% on AiShell-1 and 3.4% on Librispeech-960, comparable to conventional ANN transformers while reducing theoretical inference energy consumption by $4.64\times $ and $4.32\times $ , respectively. IML-Spikeformer marks an advance of scalable SNN architectures for large-scale speech processing in both task performance and energy efficiency. Our source code and model checkpoints are publicly available at github.com/Pooookeman/IML-Spikeformer.

Spiking Neural Networks (SNNs) provide an energy-efficient way to extract 3D spatio-temporal features. Point clouds are sparse 3D spatial data, which suggests that SNNs should be well-suited for processing them. However, when applying SNNs to point clouds, they often exhibit limited performance and fewer application scenarios. We attribute this to inappropriate preprocessing and feature extraction methods. To address this issue, we first introduce the Spike Voxel Coding (SVC) scheme, which encodes the 3D point clouds into a sparse spike train space, reducing the storage requirements and saving time on point cloud preprocessing. Then, we propose a Spike Sparse Convolution (SSC) model for efficiently extracting 3D sparse point cloud features. Combining SVC and SSC, we design an efficient 3D SNN backbone (E-3DSNN), which is friendly with neuromorphic hardware. For instance, SSC can be implemented on neuromorphic chips with only minor modifications to the addressing function of vanilla spike convolution. Experiments on ModelNet40, KITTI, and Semantic KITTI datasets demonstrate that E-3DSNN achieves state-of-the-art (SOTA) results with remarkable efficiency. Notably, our E-3DSNN (1.87M) obtained 91.7% top-1 accuracy on ModelNet40, surpassing the current best SNN baselines (14.3M) by 3.0%. To our best knowledge, it is the first direct training 3D SNN backbone that can simultaneously handle various 3D computer vision tasks (e.g., classification, detection, and segmentation) with an event-driven nature.

Kexin Wang, Yuhong Chou, Di Shang, Shijie Mei, Jiahong Zhang, Yanbin Huang, Man Yao, Bo Xu, Guoqi Li. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Linear attention mechanisms deliver significant advantages for Large Language Models (LLMs) by providing linear computational complexity, enabling efficient processing of ultra-long sequences (e.g., 1M context). However, existing Sequence Parallelism (SP) methods, essential for distributing these workloads across devices, become the primary performance bottleneck due to substantial communication overhead. In this paper, we introduce ZeCO (Zero Communication Overhead) sequence parallelism for linear attention models, a new SP method designed to overcome these limitations and achieve practically end-to-end near-linear scalability for long sequence training. For example, training a model with a 1M sequence length across 64 devices using ZeCO takes roughly the same time as training with an 16k sequence on a single device. At the heart of ZeCO lies All-Scan, a novel collective communication primitive. All-Scan provides each SP rank with precisely the initial operator state it requires while maintaining a minimal communication footprint, effectively eliminating communication overhead. Theoretically, we prove the optimaity of ZeCO, showing that it introduces only negligible time and space overhead. Empirically, we compare the communication costs of different sequence parallelism strategies and demonstrate that All-Scan achieves the fastest communication in SP scenarios. Specifically, on 256 GPUs with an 8M sequence length, ZeCO achieves a 60\% speedup compared to the current state-of-the-art (SOTA) SP method. We believe ZeCO establishes a clear path toward efficiently training next-generation LLMs on previously intractable sequence lengths.

The ambition of brain-inspired Spiking Neural Networks (SNNs) is to become a low-power alternative to traditional Artificial Neural Networks (ANNs). This work addresses two major challenges in realizing this vision: the performance gap between SNNs and ANNs, and the high training costs of SNNs. We identify intrinsic flaws in spiking neurons caused by binary firing mechanisms and propose a Spike Firing Approximation (SFA) method using integer training and spike-driven inference. This optimizes the spike firing pattern of spiking neurons, enhancing efficient training, reducing power consumption, improving performance, enabling easier scaling, and better utilizing neuromorphic chips. We also develop an efficient spike-driven Transformer architecture and a spike-masked autoencoder to prevent performance degradation during SNN scaling. On ImageNet-1k, we achieve state-of-the-art top-1 accuracy of 78.5%, 79.8%, 84.0%, and 86.2% with models containing 10 M, 19 M, 83 M, and 173 M parameters, respectively. For instance, the 10 M model outperforms the best existing SNN by 7.2% on ImageNet, with training time acceleration and inference energy efficiency improved by 4.5× and 3.9×, respectively. We validate the effectiveness and efficiency of the proposed method across various tasks, including object detection, semantic segmentation, and neuromorphic vision tasks. This work enables SNNs to match ANN performance while maintaining the low-power advantage, marking a significant step towards SNNs as a general visual backbone.

Spiking neural networks (SNNs) are emerging as an energy-efficient alternative to traditional artificial neural networks (ANNs) due to their unique spike-based event-driven nature. Coding is crucial in SNNs as it converts external input stimuli into spatio-temporal feature sequences. However, most existing deep SNNs rely on direct coding that generates powerless spike representation and lacks the temporal dynamics inherent in human vision. Hence, we introduce Gated Attention Coding (GAC), a plug-and-play module that leverages the multi-dimensional gated attention unit to efficiently encode inputs into powerful representations before feeding them into the SNN architecture. GAC functions as a preprocessing layer that does not disrupt the spike-driven nature of the SNN, making it amenable to efficient neuromorphic hardware implementation with minimal modifications. Through an observer model theoretical analysis, we demonstrate GAC's attention mechanism improves temporal dynamics and coding efficiency. Experiments on CIFAR10/100 and ImageNet datasets demonstrate that GAC achieves state-of-the-art accuracy with remarkable efficiency. Notably, we improve top-1 accuracy by 3.10% on CIFAR100 with only 6-time steps and 1.07% on ImageNet while reducing energy usage to 66.9% of the previous works. To our best knowledge, it is the first time to explore the attention-based dynamic coding scheme in deep SNNs, with exceptional effectiveness and efficiency on large-scale datasets. Code is available at https://github.com/bollossom/GAC.

Various linear complexity models, such as Linear Transformer (LinFormer), State Space Model (SSM), and Linear RNN (LinRNN), have been proposed to replace the conventional softmax attention in Transformer structures. However, the optimal design of these linear models is still an open question. In this work, we attempt to answer this question by finding the best linear approximation to softmax attention from a theoretical perspective. We start by unifying existing linear complexity models as the linear attention form and then identify three conditions for the optimal linear attention design: (1) Dynamic memory ability; (2) Static approximation ability; (3) Least parameter approximation. We find that none of the current linear models meet all three conditions, resulting in suboptimal performance. Instead, we propose Meta Linear Attention (MetaLA) as a solution that satisfies these conditions. Our experiments on Multi-Query Associative Recall (MQAR) task, language modeling, image classification, and Long-Range Arena (LRA) benchmark demonstrate that MetaLA is more effective than the existing linear models.

Spiking neural networks (SNNs) are brain-inspired energy-efficient models that encode information in spatiotemporal dynamics. Recently, deep SNNs trained directly have shown great success in achieving high performance on classification tasks with very few time steps. However, how to design a directly-trained SNN for the regression task of object detection still remains a challenging problem. To address this problem, we propose EMS-YOLO, a novel directly-trained SNN framework for object detection, which is the first trial to train a deep SNN with surrogate gradients for object detection rather than ANN-SNN conversion strategies. Specifically, we design a full-spike residual block, EMS-ResNet, which can effectively extend the depth of the directly-trained SNN with low power consumption. Furthermore, we theoretically analyze and prove the EMS-ResNet could avoid gradient vanishing or exploding. The results demonstrate that our approach outperforms the state-of-the-art ANN-SNN conversion methods (at least 500 time steps) in extremely fewer time steps (only 4 time steps). It is shown that our model could achieve comparable performance to the ANN with the same architecture while consuming 5.83× less energy on the frame-based COCO Dataset and the event-based Gen1 Dataset. Our code is available in https://github.com/BICLab/EMS-YOLO.