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
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:2608.30720v1 Announce Type: new
Abstract: Representational similarity is foundational to analyses of deep networks, yet distances between point-valued representations are not intrinsically tied...
By Kieran Murphy
The paper investigates how many transformer components influence a token prediction by measuring the absolute contribution of each unit and channel to the logit. It finds that thousands of components contribute to a single prediction, yet a small subset—often just dozens—carries the majority of the predictive mass. Across models ranging from 124 M to 7 B parameters, the proportion of the model involved in a prediction remains around one to three percent, independent of size, and the study demonstrates that specific components can be directly read and written to modify model behavior without additional training.
By Mark Oskin
arXiv:2607. 01218v1 Announce Type: cross Abstract: Transformers use the same forward computation stream to both predict the next token and store useful state for future token predictions.
By Giovanni Monea, Nathan Godey, Kiant\'e Brantley, Yoav Artzi
arXiv:2607. 18363v1 Announce Type: cross Abstract: Feed-forward networks hold two thirds of a transformer's non-embedding parameters, yet the architecture has not received a necessity test that controls parameters, compute, and depth at once.
By Henry Ndubuaku, Karen Mosoyan, Jakub Mroz, Noah Cylich, Satyajit Kumar, Parkirat Sandhu, Roman Shemet, Justin H Lee
arXiv:2606. 02559v1 Announce Type: cross Abstract: Post-training compression of Large Language Models (LLMs) removes entire architectural components, either deleting them or replacing them with fitted modules.
By Elia Cunegatti, Marcus Vukojevic, Erik Nielsen, Giovanni Iacca
arXiv:2606. 26538v1 Announce Type: cross Abstract: Deep Transformers are composed of uniformly stacked residual blocks, yet their deepest layers often add little value.
By Huzama Ahmad, Cao Viet Hai Nam, Se-Young Yun
arXiv:2607. 08946v1 Announce Type: new Abstract: A transformer can be built from operators that are legible by construction -- bounded, named units that read as fuzzy set operations rather than dense activations -- but legibility must be pressed for during training, and the pressure has a failure mode.
By Mark Oskin
arXiv:2602.06471v2 Announce Type: replace
Abstract: The architectural shape of dense Transformers has remained remarkably stable: narrow-wide-narrow feed-forward networks (FFNs) consume most non-embe...
By Feng-Ting Liao, Guan-Ting Yi, Tzu-Quan Lin, Meng-Hsi Chen, Da-shan Shiu
arXiv:2607. 02964v1 Announce Type: cross Abstract: A central goal of mechanistic interpretability is to understand how neural networks work and what each individual component does.
By Arnau Marin-Llobet, Stefan Heimersheim
arXiv:2606. 27449v1 Announce Type: new Abstract: Multi-head attention conventionally partitions the hidden dimension equally across all heads at every layer, enforcing an identical representational subspace dimension (dh = dmodel/h) throughout the models depth.
By Shubham Aggarwal