The paper proposes a principled way to design hybrid transformer architectures that combine Full Attention (FA) and Linear Attention (LA). By introducing two intervention metrics—RoPE Frequency Importance Score (RFIS) and RoPE Positional Dependence (RPD)—the authors identify a clear taxonomy of retrieval and positional heads, defining a Global Positional Band (GPBand) that aligns with training-length positional scales. Using these insights, they build a Head‑wise Hybrid Architecture (HwH) that assigns FA to global retrieval and LA to local positional modeling, achieving strong language modeling, improved retrieval, and superior zero‑shot long‑context extrapolation compared to standard Transformers and other hybrids.
By Runlin Shi, Bojian Yin, Guoqi Li
Positional encodings (PEs) are essential for Transformers. Yet designing effective PEs for non-Euclidean graphs remains challenging.
arXiv:2606. 25293v1 Announce Type: new Abstract: Positional encodings (PEs) are essential for Transformers.
By Yipeng Zhang, Zhongtian Sun, Pietro Li\`o, Kelin Xia
arXiv:2602. 18948v2 Announce Type: replace Abstract: Transformer models contain substantial internal redundancy arising from coordinate-dependent representations and continuous symmetries, in model space and in head space, respectively.
By J. Fran\c{c}ois, L. Ravera
arXiv:2609.37921v1 Announce Type: new
Abstract: What are the inductive biases of a Transformer architecture? Existing theory on how the forward pass shapes representations either considers whether Tr...
By Erkan Turan, Gaspard Abel, Maks Ovsjanikov
arXiv:2609.15975v1 Announce Type: cross
Abstract: Transformer representations evolve through learned additive transformations that either preserve their current direction or redirect it. We study thi...
By Shwai He, Haichao Zhang, Shen Yan
arXiv:2509. 22259v4 Announce Type: replace-cross Abstract: We study the extent to which rotary position encodings (RoPE), a recent transformer position encoding algorithm broadly adopted in large language models (LLMs) and vision transformers (ViTs), can be applied to graph-structured data.
By Isaac Reid, Arijit Sehanobish, Cederik H\"ofs, Bruno Mlodozeniec, Leonhard Vulpius, Federico Barbero, Adrian Weller, Krzysztof Choromanski, Richard E. Turner, Petar Veli\v{c}kovi\'c
arXiv:2609.38109v1 Announce Type: cross
Abstract: The attention operation is naively position invariant. However, positional information is fundamental to natural language, and therefore a variety of...
By Cutter Dawes, Nick Alonso, Tom Figliolia, Beren Millidge
arXiv:2510. 24342v2 Announce Type: replace Abstract: Prior brain-AI alignment studies are typically constrained by specific inputs and tasks, limiting their ability to capture organizational properties across models with different modalities.
By Silin Chen, Yuzhong Chen, Caiwei Wang, Zifan Wang, Junhao Wang, Zifeng Jia, Keith M Kendrick, Tuo Zhang, Lin Zhao, Dezhong Yao, Tianming Liu, Xi Jiang
arXiv:2609.01231v1 Announce Type: new
Abstract: We prove that a multi-head scaled dot product attention can be viewed as a parameter identification strategy. The ratio of unidentified parameters to t...
By W. Ross Morrow
arXiv:2606. 13276v1 Announce Type: cross Abstract: Weight-space geometry plays a central role in neural network optimization, yet manifold constraints are often applied uniformly across all weight matrices.
By Kirato Yoshihara
The paper examines how to allocate attention heads and head dimensions across Transformer layers to balance expressivity and efficiency. It provides a mathematical analysis of early layers’ role in information extraction and characterizes the trade‑off between head count and dimension under a fixed parameter budget. The authors prove a saturation effect of softmax activations, showing that increasing head dimensions yields diminishing returns, especially for long sequences, and propose strategies for efficient parameter allocation across layers.
By Ruoxi Yu, Haotian Jiang, Jingpu Cheng, Penghao Yu, Qianxiao Li, Zhong Li