The paper introduces Successive Capacity Growth (SCG), a method for adaptively expanding Vision Transformer encoders in Joint-Embedding Predictive Architectures (JEPAs). SCG starts with a minimal encoder and incrementally increases width or depth based on a task‑agnostic test‑and‑verify mechanism, while a Sketched Isotropic Gaussian Regularizer (SIGReg) keeps learned semantic dimensions independent. Experiments on multi‑object dynamics and 2D navigation tasks show that SCG achieves up to 20.3% better prediction loss than fixed small baselines and 23% better than fixed large models, with far greater parameter efficiency and no false‑positive expansions.
By Frederik Berenz
arXiv:2605.12491v2 Announce Type: replace
Abstract: Vision Transformers (ViTs) learn rich visual-semantic representations through all-to-all self-attention among patch tokens. However, this design im...
By Alan Z. Song, Yinjie Chen, Mu Nan, Deva Ramanan, Michael J. Tarr, Andrew F. Luo
arXiv:2606. 16112v1 Announce Type: cross Abstract: Residual architectures are ubiquitous in deep learning, but they suffer from a subtle structural limitation: the norm of the residual stream can grow rapidly with depth.
By Tom\'as Figliolia, Beren Millidge
arXiv:2607. 16314v1 Announce Type: cross Abstract: World models, especially based on JEPA architectures, have been shown to learn robust dynamics of various environments.
By Usman M. Khan
Vision-language models commonly project all tokens produced by a pretrained vision encoder into a large language model. However, final-layer features can discard text, local attributes, and spatial relationships, while high-resolution inputs substantially increase context length and inference latency.
arXiv:2606. 09936v1 Announce Type: cross Abstract: World models are now built on substantially different computational substrates.
By Bhavith Chandra Challagundla, Sanskar Pandey, Param Thakkar, Rishikesh Mallagundla, Yugandhar Reddy Gogireddy, Wenhao Lu, Hindol Roy Choudhury, Shravani Challagundla, Mohamed Deraz Nasr, Spursh Deshpande
arXiv:2610.28074v1 Announce Type: new
Abstract: As a prototypical latent world model, Joint-Embedding Predictive Architectures (JEPAs) provide strong semantic representations. However, dense patch re...
By Peisong Wen, Qianqian Xu, Minye Lei, Siran Dai, Yang Liu, Qingming Huang
The paper introduces DARTS, a method for tuning decoder representations during model merging. It addresses representation bias in autoregressive decoders by using an entropy‑weighted L1 loss and a per‑position additive bias to correct errors that accumulate across token positions. Experiments on code generation, mathematical reasoning, and instruction following with Llama‑2‑7B show that DARTS improves performance over standard surgery while adding only 0.1% extra parameters.
By Aaryan Ajay Sharma, Sai Nishanth Padala, Seganrasan Subramanian
The paper introduces Action-Contrastive Masked Transition Modeling (AC‑MTM), a method that stabilizes Joint‑Embedding Predictive Architectures (JEPAs) without relying on Gaussian regularization. AC‑MTM adds a training‑only inverse‑dynamics head that uses Action‑NCE to force each latent transition to identify its generating action, thereby preventing encoder collapse. Experiments on pixel‑control and multi‑object visual tasks show that AC‑MTM trains stably from scratch and matches or surpasses the performance of SIGReg, achieving up to a 24‑point improvement on the OGBench Visual Scene benchmark.
LoopLook introduces a lightweight encoder-side recurrent perception framework that equips pretrained Vision‑Language Models with adaptive visual depth. By repeatedly revisiting a middle‑to‑late window of the vision encoder and injecting question representations at each loop, it refines visual evidence in a task‑directed manner. The learned stopping mechanism and token‑wise Mixture‑of‑Looks allow each visual token to selectively integrate information across loops, yielding consistent performance gains across multiple backbones and benchmarks.
By Wencheng Ye, Anning Hu, Hongye Fang, Tianyi Wang, Junchi Yan
arXiv:2605. 18848v3 Announce Type: replace Abstract: This paper introduces Exact Linear Attention (ELA), a mechanism that achieves linear computational complexity for Transformer attention by exploiting the exact decomposition property of kernel functions, thereby eliminating approximation error.
By Weinuo Ou
The paper investigates how to scale a Transformer-based vision‑language model trained from scratch on multimodal inputs under a fixed compute budget. It finds that minimal loss follows a predictable compute law while optimal model size and token count obey power‑law scaling, with language and multimodal objectives showing distinct allocation trends. The study also proposes a budget‑compensation rule that allows additional multimodal tokens to offset text‑efficiency penalties, and demonstrates that native multimodal pre‑training improves spatial reasoning and few‑shot learning.
By Haoyuan Wu, Aoqi Wu, Hai Wang, Jiajia Wu, Jinxiang Ou, Bei Yu