The transition to Data Processing Unit (DPU)-centric architectures has become the de-facto standard in modern cloud networks, enabling infrastructure offload and improved host resource utilization. However, the fixed hardware limits of DPUs increasingly fail to keep pace with the rapid growth of host compute density and network-intensive workloads. As a result, when DPU resources are saturated, host compute capacity often remains underutilized due to insufficient network provisioning.

The hypothesis of \textit{Universality} in interpretability suggests that different neural networks may converge to implement similar algorithms on similar tasks. In this work, we investigate two mainstream architectures for language modeling, namely Transformers and Mambas, to explore the extent of their mechanistic similarity. We propose to use Sparse Autoencoders (SAEs) to isolate interpretable features from these models and show that most features are similar in these two models. We also validate the correlation between feature similarity and~\univ. We then delve into the circuit-level analysis of Mamba models and find that the induction circuits in Mamba are structurally analogous to those in Transformers. We also identify a nuanced difference we call \emph{Off-by-One motif}: The information of one token is written into the SSM state in its next position. Whilst interaction between tokens in Transformers does not exhibit such trend.

Synthetic aperture passive positioning (SAPP) has gradually become a research hotspot in the field of single-station passive positioning. Except for the acoustic band, almost all research in the microwave band only focuses on the spaceborne scenarios. In this paper, we attempt to extend SAPP to more scenarios, such as airborne and vehicle-mounted applications, and introduce the motion error problem. The difficulty is that, different from the problem of SAR motion compensation, the target area of passive positioning is unknown, and the compensation matrix cannot be calculated directly. We propose an iterative motion compensation (IMC) algorithm to solve it. Both simulation and vehicle experiment results verify the effectiveness of the IMC algorithm. The results show that IMC improves the positioning accuracy by two orders of magnitude in the presence of motion error and exhibits strong convergence.

Siyu Yuan, Jiangjie Chen, Ziquan Fu, Xuyang Ge, Soham Shah, Charles Jankowski, Yanghua Xiao, Deqing Yang. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.