RT-SEMamba is a fully causal speech enhancement model that uses causal time‑frequency Mamba blocks instead of Transformer‑based architectures, allowing efficient long‑form inference with a fixed‑size recurrent state. The authors introduce a progressive knowledge distillation strategy that compresses an 8‑layer teacher into a single‑layer student by jointly distilling spectral outputs and intermediate representations. On the Voicebank‑DEMAND benchmark, the 8‑layer model achieves 3.32 PESQ under a 25 ms latency constraint, while the distilled 1‑layer student improves from 3.06 to 3.18 PESQ, maintains the same steady‑state real‑time factor, and runs 2.64× faster than the teacher.
By Rong Chao, Sung-Feng Huang, Moreno La Quatra, Sabato Marco Siniscalchi, Wen-Huang Cheng, Szu-Wei Fu, Yu Tsao
arXiv:2603. 15590v2 Announce Type: replace Abstract: There have been numerous attempts to distill quadratic attention-based large language models (LLMs) into sub-quadratic linearized architectures.
By Lukas Hauzenberger, Niklas Schmidinger, Thomas Schmied, Anamaria-Roberta Hartl, David Stap, Pieter-Jan Hoedt, Maximilian Beck, Sebastian B\"ock, G\"unter Klambauer, Sepp Hochreiter
arXiv:2606. 16429v1 Announce Type: new Abstract: Hybrid linear attention models offer an appealing path to faster long-context inference: they reduce the quadratic cost and KV-cache burden of full softmax attention while retaining much of the quality of Transformer models.
By Zhongzhu Zhou, Qingyang Wu, Junxiong Wang, Mayank Mishra, Shuaiwen Leon Song, Ben Athiwaratkun, Chenfeng Xu
arXiv:2607. 06796v1 Announce Type: cross Abstract: Deep learning has achieved remarkable success in various domains including time series analysis, computer vision and natural language processing.
By Javidan Abdullayev, Maxime Devanne, Jonathan Weber, Germain Forestier
The paper investigates how attention dynamics evolve across recurrent depth in language models, finding that attention support stabilizes early while hidden states and outputs take longer. It proposes WISE, a training‑free method that uses full attention in early steps and then reuses the discovered sparse working set for later steps, preserving performance on multi‑hop QA tasks. Experiments show that WISE maintains quality up to 2K context, offers measurable speedups, and highlights the importance of recurrent discovery of attention support.
By Ke Wan, Chen Chen
arXiv:2607. 08013v1 Announce Type: new Abstract: Federated Learning (FL) empowers multiple clients to collaboratively learn a model, enlarging the training data of each client for high accuracy while protecting data privacy.
By Shuo Huai, Di Liu, Hao Kong, Xiangzhong Luo, Weichen Liu, Ravi Subramaniam, Christian Makaya, Qian Lin
Representation alignment has become an effective way to accelerate diffusion transformer training and improve generation quality. Recent self-alignment methods, such as SRA and Self-Flow, further remove the dependency on external pretrained encoders by constructing alignment within the diffusion model itself.
FeatureFormer is a neural performance predictor that adds explicit node-wise encodings of FLOPs, parameter counts, and memory proxies to a gated graph attention architecture. It is designed to improve latency and energy prediction for neural networks on edge devices, addressing the limitation of existing GNN and transformer predictors that largely ignore node-level computational cost. The authors also introduce NNEQ, a large-scale energy consumption dataset, and show through extensive experiments that FeatureFormer achieves state‑of‑the‑art performance across both metrics, including challenging out‑of‑domain settings, while the encoding can broadly enhance existing predictors with negligible overhead.
By Matthew Grenier, William Hammer, Andrew Heuer, Nikhil Krishna, Yi Wang, Ramtin Zand
StreamTTT is a streaming vision-language model that balances real-time perception with long-term memory by writing long-range history into fast weights outside the attention context, while keeping a short sliding key-value cache for recent evidence. The model is trained on both offline long-video QA and a new real-time QA corpus, and it outperforms SimpleStream-4B on OVO-Bench by 1.4 points in real-time perception and 3.7 points in backward tracing. StreamTTT-4B also competes with the larger SimpleStream-8B on the StreamingBench Real-Time Visual Understanding subset.
By Joya Chen, Zeyun Zhong, Mike Zheng Shou
With the widespread deployment of basic models in edge intelligence, communication bandwidth has become a core bottleneck restricting the scalability of federated learning. Although one-shot federated learning alleviates this problem by minimizing communication rounds, existing iterative fine-tuning or knowledge distillation methods still face challenges such as high server-side computational costs and hyperparameter sensitivity.
The paper introduces sLoTh, a parameter‑efficient continual learning framework for sparse event‑based vision transformers. By freezing the backbone and limiting plasticity to low‑rank attention updates (seLoRA) and shared neuronal threshold modulation, sLoTh adapts to new tasks while updating less than 1% of the parameters and avoiding replay buffers. Experiments on CIFAR‑100, Tiny‑ImageNet, ImageNet‑100, and ImageNet‑R show competitive rehearsal‑free performance across up to 100 tasks and achieve roughly 6.5× lower energy consumption than dense vision transformers.
arXiv:2607. 13395v1 Announce Type: new Abstract: The pursuit of autonomously self-improving models has attracted growing interest in the era of large-scale foundation models.
By Jing-Xiao Liao, Tianwei Zhang, Yu-Hao Jiang, Feifei Zhang, Hang-Cheng Dong, Feng-Lei Fan