The paper introduces LLMAE, a technique that transforms a pretrained decoder-only language model into a continuous text autoencoder by inserting a fixed-length latent bottleneck into its internal activations. Using a 270M Gemma 3 model with structured attention masks, LoRA adaptation, and KL regularization, LLMAE achieves near-perfect reconstruction of text sequences up to 1024 tokens. The authors further show that the resulting latent representation can be leveraged to train a latent text diffusion model for detailed image captioning, demonstrating downstream utility.
By Arkanath Pathak, Unnat Jain, Alexander C. Berg
arXiv:2607. 27372v1 Announce Type: new Abstract: The deep learning revolution, kicked off by AlexNet, taught us that end-to-end training beats decomposing a problem into hand-designed stages.
By Alexi Gladstone, Heng Ji, Yilun Du
arXiv:2606. 16408v1 Announce Type: new Abstract: We introduce MUNI, an end-to-end multimodal latent diffusion framework for any-to-any generation that unifies subset-conditioned cross-modal generation and unconditional joint sampling through a shared stochastic latent.
By Kyeongmin Yeo, Yunhong Min, Minhyuk Sung
arXiv:2610.01942v1 Announce Type: new
Abstract: Predicting the future evolution of a scene is a fundamental capability for world modeling. Recent work has shown that operating in the feature space of...
By Efstathios Karypidis, Spyros Gidaris, Nikos Komodakis
The paper introduces a post‑training approach that enables a single inference process to transition from text reasoning to image synthesis, eliminating the need for explicit modality switching. Using the 14B BAGEL model, the authors demonstrate that targeted post‑training data and reward‑weighted training improve multimodal image generation across four independent T2I benchmarks. The study highlights the benefits of joint text‑image generation and strategic data selection for enhancing T2I performance.
By Jiahui Chen, Philippe Hansen-Estruch, Xiaochuang Han, Yushi Hu, Emily Dinan, Amita Kamath, Michal Drozdzal, Reyhane Askari-Hemmat, Luke Zettlemoyer, Marjan Ghazvininejad
arXiv:2112.09646v2 Announce Type: replace
Abstract: The generation of discontinuous distributions is a difficult task for most known frameworks such as generative autoencoders and generative adversar...
By Mariia Drozdova, Vitaliy Kinakh, Guillaume Qu\'etant, Tobias Golling, Slava Voloshynovskiy
MUNITE is a latent‑variable framework that unifies encoding and latent generation for any‑to‑any multimodal generation. By treating different amounts of observed evidence as varying inference problems, it models the conditional distribution over a shared latent representation, using a single conditional flow model. The approach employs self‑distillation with conditional flow matching to learn from incomplete data, achieving competitive or superior generation quality and coherence across datasets such as PolyMNIST‑D‑Q, FFHQ64, and image‑text‑audio.
By Kyeongmin Yeo, Minhyuk Sung
arXiv:2609.37775v1 Announce Type: cross
Abstract: Pretrained visual representations support image generation, but may not fully preserve the fine-grained details needed for faithful reconstruction. M...
By Xuanyu Zhu, Yan Bai, Yang Shi, Yihang Lou, Yuanxing Zhang, Tengfei Liu, Jing Jin, Yuan Zhou
arXiv:2603.20111v2 Announce Type: replace
Abstract: The Joint-Embedding Predictive Architecture (JEPA) is often seen as a non-generative alternative to likelihood-based self-supervised learning, emph...
By Moritz G\"ogl, Christopher Yau
The paper introduces SCION, a single-model framework that jointly learns generative and self‑supervised representation tasks without relying on labels or pretrained encoders. SCION uses a pixel‑space encoder conditioned on a flow timestep and a shared embedding, training contrastive and flow‑matching objectives simultaneously with gradient‑norm balancing and stop‑gradient techniques. On ImageNet 256×256, SCION achieves state‑of‑the‑art FID scores (8.92 with JiT‑B and 5.89/3.47 with JiT‑L) surpassing prior class‑unconditional and pretrained‑conditioned methods.
By Vicente Balmaseda, Ching-Long Lin, Tianbao Yang
arXiv:2609.24088v2 Announce Type: replace
Abstract: In latent generative models, reconstruction quality is often assumed to correlate with generative performance. However, reconstruction FID (rFID) c...
By Xianghong Fang, Wenjie Shu, Tongda Xu, Wenlong Mou, Dehan Kong, Tim G. J. Rudner
The paper investigates how long, richly detailed prompts cause modern text-to-image models to lose diversity, even when many visual aspects are unspecified. It introduces PromptMoG, a training‑free method that samples prompt embeddings from a Mixture‑of‑Gaussians distribution to restore diversity while preserving semantic fidelity. The authors also present LPD‑Bench, a benchmark for evaluating fidelity and diversity under long, semantically dense prompts, and demonstrate PromptMoG’s effectiveness on four large diffusion models.
By Bo-Kai Ruan, Teng-Fang Hsiao, Ling Lo, Yi-Lun Wu, Hong-Han Shuai