The paper revisits the interpretation of depth profiles in Diffusion Transformer backbones, showing that the commonly observed inverted‑U curve reflects the influence of the prediction target (e.g., ε, u, x0) rather than a simple shift from coarse to fine semantic granularity. It argues that the semantically meaningful signal is merely an intermediate by‑product of predicting the target, and that different training setups—particularly the choice of prediction target—predictably alter the curve’s shape. The authors demonstrate how this new understanding can guide the optimal placement of REPA modules and provide code for further exploration.
By Benyuan Meng, Longtao Huang, Yang Liu, Liang Lin
arXiv:2607. 13491v1 Announce Type: cross Abstract: Looped Transformers scale sequential computation by applying a compact stack of physical blocks for multiple rounds, increasing unrolled depth without increasing stored parameters.
By Shuzhen Li, Yifan Zhang, Jiacheng Guo, Quanquan Gu, Mengdi Wang
arXiv:2605. 20708v2 Announce Type: replace-cross Abstract: Diffusion Transformers (DiTs) have become a de facto backbone of modern visual generation, and nearly every major axis of their design -- tokenization, attention, conditioning, objectives, and latent autoencoders -- has been extensively revisited.
By Chao Xu, Maohua Li, Qirui Li, Yixuan Xu, Yanke Zhou, Yunhe Li, Cuifeng Shen, Hanlin Tang, Kan Liu, Tao Lan, Lin Qu, Shao-Qun Zhang
The paper presents a theoretical framework for understanding generalization and scaling in Mixture-of-Experts (MoE) Transformers. It derives a sup‑norm covering‑number bound that separates active per‑input capacity from routing combinatorics, leading to a generalization bound under a manifold data model. Additionally, a constructive approximation theorem shows how error can be reduced by scaling active capacity or increasing the number of experts, from which neural scaling laws and compute‑optimal tradeoffs are derived.
By Mansour Zoubeirou a Mayaki
arXiv:2608. 02064v1 Announce Type: new Abstract: Feed-forward networks (FFNs) account for a large fraction of Transformer parameters, yet their hidden width is usually constant across depth.
By Timur Mudarisov, Mikhail Burtsev, Radu State
arXiv:2606. 18524v1 Announce Type: new Abstract: Looped (weight-tied) Transformers apply a shared residual block $N$ times ($h \leftarrow h + \varepsilon\,f(h)$, same $f$ at each step), increasing effective depth without adding parameters.
By Shaowen Wang, Bingrui Li, Ge Zhang, Wenhao Huang, Shen Yan, Jian Li
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. 24665v1 Announce Type: cross Abstract: Modern large language models scale successfully by pairing capacity growth with efficiency, keeping per-token and deployment costs under control as capacity grows.
By Yanhao Jia, Jiepeng Wang, Haibin Huang, Chi Zhang, Erik Cambria, Xuelong Li
The paper introduces the first controlled benchmark for optimizers in discrete diffusion models, evaluating seven optimizers (AdamW, Lion, Muon, SOAP, MARS, MARS‑M, Schedule‑Free) across four diffusion formulations: masked diffusion on text8, uniform diffusion on QM9 and LM1B, and Gaussian diffusion on CelebA‑64. Each optimizer undergoes the same search protocol and is retrained with full budget and multiple seeds, revealing that AdamW, while strong, is not universally optimal and that optimizers validated on autoregressive language models (Muon, MARS‑M, SOAP) can outperform tuned AdamW on certain tasks.
By Arman Bolatov, Egor Shulgin, David Li, Abduragim Shtanchaev, Sebastian U. Stich, Maxim Panov, Eric Moulines, Peter Richt\'arik, Martin Tak\'a\v{c}
arXiv:2607. 14427v1 Announce Type: new Abstract: A depth-recurrent transformer applies a weight-tied core a variable number of times, and prior work has shown that training with a randomized recursion count yields one checkpoint usable across a range of inference depths.
By Joe Logan
arXiv:2610. 01172v1 Announce Type: new Abstract: We study learning rate (LR) scaling for hybrid architectures combining Transformer and State-Space Model (SSM) blocks, a class adopted by several recent production language models.
By Jimin Seo, Gyubok Lee, Yeonsik Jo, Kiwoong Yoo, Yeongoon Kim, Minhae Oh, Jin Woo Koo, Suhwan Kim, Nakyung Lee, Minsik Seol, Idris Nechnech, Jaehyeon Kim, Giho Lee, Jungwoo Lee
arXiv:2610.26030v2 Announce Type: replace
Abstract: Looped transformers reuse a shared block of layers to increase computational depth without increasing the number of trainable parameters. They are...
By Xuyuan Xiong, Siyu Chen, Zihan Shao, Yue Su, Weiyang Jin, Ping Luo, Zhuoran Yang, Xihui Liu