The paper presents a spiking neural network (SNN) approach that uses time-to-first-spike (TTFS) coding to limit each neuron to at most one spike per time window, enabling energy-efficient large language models (LLMs). A reference-based strategy is introduced to encode the four core LLM components—embedding layers, layer normalization, attention-related operations, and dropout—allowing the construction of a fully TTFS-based SNN architecture trained end-to-end. Experiments on BERT and GPT-2 show performance comparable to artificial neural network (ANN) counterparts on natural language understanding and common-sense reasoning, while achieving a 1.5‑billion‑parameter spiking LLM and providing an estimate of spike-related energy consumption.
By Zhuoya Zhao, Parsa Omidi, Aref Jafari, Richard Naud
The paper introduces a hardware-software co‑design framework that compresses Mixture‑of‑Experts (MoE) model weights into low‑precision, hardware‑native sparse representations, enabling efficient execution on Sparse Tensor Cores (SpTCs). By relaxing discrete support selection through continuous reparameterization, the method jointly optimizes quantized weights and supports a router‑weighted reconstruction objective, achieving up to 4.35 percentage‑point gains in joint sparse‑quantization accuracy while retaining 96.09% of the original model’s performance. A custom grouped sparse GEMM kernel further boosts inference speed, outperforming NVIDIA’s baseline by up to 1.65× and reducing latency by up to 4.03× on B200 GPUs.
By Kwanhee Lee, Namhoon Lee, Dan Alistarh
arXiv:2603. 29002v3 Announce Type: replace-cross Abstract: Modern large language models (LLMs) increasingly depends on efficient long-context processing and generation mechanisms, including sparse attention, retrieval-augmented generation (RAG), and compressed contextual memory, to support complex reasoning.
By Zifan He, Rui Ma, Yizhou Sun, Jason Cong
arXiv:2608. 15602v1 Announce Type: cross Abstract: While binary quantization theoretically promises extreme compression and acceleration for Large Language Models (LLMs), existing research often overlooks the necessity of specialized hardware kernels, thus failing to unleash the full acceleration potential due to persistent reliance on expensive floating-point arithmetic or runtime dequantization overheads.
By Qingyao Yang, Runming Yang, He Xiao, Wendong Xu, Junyu Chen, Haobo Liu, Chenchen Ding, Ruihan Hu, Yik-Chung Wu, Ngai Wong
Elastic Threshold Attention (ETA) is a trainable attention mechanism that dynamically predicts contextual thresholds from query representations, enabling selective pruning of KV cache tokens during long‑context decoding. By multiplicatively suppressing sub‑threshold logits during training, ETA avoids representation collapse and eliminates localized attention sinks, allowing a 1.45B model to match dense attention performance at roughly 85% training sparsity and 38% active decode density. At inference, a custom Triton kernel achieves up to 2.5× faster decoding on sequences up to 512K tokens, and an offline calibration step can further reduce compute by 27% by freezing per‑head thresholds.
By Themistoklis Haris, Henry Li, Maryam Karimzadehgan
SparseGDN is a hardware‑aware framework that makes Gated Delta Networks (GDNs) more efficient for long‑context sequence modeling. It combines structured intra‑chunk approximation with an input‑dependent recurrent‑event selector, producing outputs for every token while reducing dense computation. The GPU implementation updates only selected events and reconstructs token outputs from compact state snapshots, cutting computation and memory traffic.
By Jiedong Zhuang, Peng Zhao, Jialin Tang, Haoji Hu
arXiv:2606. 03026v1 Announce Type: cross Abstract: Spiking language models expose activation sparsity that dense Transformer runtimes do not directly exploit.
By Ting Liu
arXiv:2606. 09079v1 Announce Type: cross Abstract: Conventional LLMs keep the full KV cache loaded during decoding, causing a severe GPU memory bottleneck for ultra-long context serving.
By Yan Wang, Qifan Zhang, Jiachen Yu, Tian Liang, Dongyang Ma, Xiang Hu, Zibo Lin, Chunyang Li, Zhichao Wang, Jia Li, Yujiu Yang, Haitao Mi, Dong Yu
arXiv:2609.38166v1 Announce Type: cross
Abstract: Recent LLMs increasingly adopt hybrid designs that replace standard attention with linear attention, such as Gated DeltaNet (GDN) and Kimi Delta Atte...
By Yi Pan, Haocheng Xi, Kan Zhu, Xingyang Li, Yibo Wu, Mayank Mishra, Hongtao Zhang, William X. Zheng, Baris Kasikci, Song Han, Kurt Keutzer, Rishabh Iyer, Ion Stoica
arXiv:2607. 08786v1 Announce Type: cross Abstract: With the growing deployment of large language models (LLMs), LLM inference cost has become a key challenge.
By Tao Lu, Haoyu Wang, Zonghui Wang, Keshen Xiang, Jiaheng Zhang, Wenzhi Chen
arXiv:2605. 01910v2 Announce Type: replace-cross Abstract: Autoregressive decoding becomes bandwidth-limited at long contexts, as generating each token requires reading all $n_k$ key and value vectors from KV cache.
By Kyle Lee, Corentin Delacour, Kevin Callahan-Coray, Kyle Jiang, Can Yaras, Samet Oymak, Tathagata Srimani, Kerem Y. Camsari
arXiv:2606. 01117v1 Announce Type: cross Abstract: Extreme multi-label classification (XMC) involves learning models over large output spaces with millions of labels, making the output layer a memory-compute bottleneck.
By Nasib Ullah, Jinbin Zhang, Jean Lucien Randrianantenaina, Erik Schultheis, Rohit Babbar