AttnFusion: gated cross-attention for fine-grained visual injection in latent diffusion models

VisReflect: Latent Visual Reflection for Fine-Grained Perception in Long Visual Context

Large Vision Language Models (LVLMs) have achieved remarkable success on vision-language tasks, yet fine-grained perception over high-resolution images and long-context videos remains challenging. As the number of visual tokens increases, the visual attention sink phenomenon becomes increasingly severe, causing irrelevant tokens to absorb a disproportionate amount of attention mass. Recent approaches attempt to mitigate this issue by explicitly predicting bounding boxes or temporal spans and re-encoding the cropped visual regions. Such methods depend on unreliable numeric localization in the discrete token space and incur significant computational overhead due to additional forward passes. In this work, we propose **VisReflect**, a simple yet effective framework that improves fine-grained perception in long visual contexts through latent visual reflection. Instead of decoding intermediate predictions into discrete tokens, the model generates continuous visual reflection that represents question-relevant visual features in the latent space. These reflections selectively emphasize salient regions or frames, guiding attention towards relevant visual tokens within a single forward pass. We conduct comprehensive evaluations on challenging high-resolution image benchmarks, including BLINK, V*, and HRBench-4K/8K, as well as video understanding benchmarks such as MVBench, VideoMME, and MLVU. Our method consistently improves over strong baselines, achieving gains of 4.1% on image benchmarks and 1.8% on video benchmarks. Compared with zooming-based methods, our model achieves comparable performance while reducing inference time by roughly 44% on video understanding.

Chimera: Designing and Chinchilla-Scaling Hybrid Visual Diffusion Transformers

Chimera, a hybrid visual diffusion backbone with a principled scaling recipe that combines Kimi Delta Attention for long-context state tracking with O(N) complexity, interleaved Multi-head Latent Attention (MLA) for direct global interaction, and modality-aware short convolutions for local spatiotemporal context is introduced.

A Hybrid CLIP-Diffusion Architecture for Transparent Deepfake Detection and Explainability Evaluation

Tiper: Tokenwise Prompt-Free Learning

Vision-language models (VLMs), such as contrastive language-image pre-training (CLIP), exhibit powerful zero-shot generalization capabilities. Parameter-efficient fine-tuning (PEFT) techniques, notably prompt learning, have been extensively explored to adapt these models to downstream tasks. However, their efficacy remains constrained when transferred to specialized domains like remote sensing. We argue that the bottleneck stems not merely from the limited parameters of prompts, but essentially from the disruption of the input’s original image–text features and the lack of deep cross-modal alignment. In particular, existing methods typically rely on global attention or coarse-grained feature mapping. This inadvertently corrupts the original input representations, thereby impairing the model’s inherent generalization. Furthermore, their isolated unimodal gradient updates fail to bridge the semantic gap inherent in complex remote sensing scenes. To address these challenges, we propose tokenwise prompt-free learning (Tiper), shifting the optimization paradigm from introducing external prompts to precisely recalibrating the critical tokens that govern classification outputs. In particular, Tiper employs a hierarchical learner to supersede global prompts. Crucially, this learner intervenes exclusively on the specific core tokens (i.e., the CLS token in the visual branch and the EOT token in the textual branch), leaving other original input representations unperturbed. This fine-grained strategy effectively balances domain adaptation with the preservation of inherent generalization. Finally, we design the learner as a cross-modal coupled bridge with shared weights, enabling it to synchronously receive gradient feedback from both modalities and fostering profound multimodal collaboration. Extensive experiments validate our method on eight public remote sensing datasets covering diverse scenes and resolutions. In the base-to-new generalization task, Tiper outperforms the strong baseline MaPLe with a significant 3.7% improvement in the harmonic mean (HM). Notably, without relying on any external large-scale domain models, Tiper surpasses the latest domain-specific prompt learning methods (e.g., domain-controlled prompt learning (DCPL), domain prompt learning with quaternion networks (DPLQ)), demonstrating its superior adaptability for remote sensing image scene classification.

Zero-shot cross-domain image composition via self attention injection.

Leveraging the robust generative priors of diffusion models, image composition has achieved remarkable progress. However, existing approaches continue to grapple with a persistent dilemma: the trade-off between maintaining the structural fidelity of the source object and achieving deep stylistic harmonization with the background. We attribute this limitation to two primary factors: 1) the insufficient disentanglement of geometric structure and visual appearance in current architectures, leading to conflicts during the generation process; and 2) the reliance on global statistical alignment techniques, which merely adjust tonal distributions but fail to capture complex semantic stylistic patterns. To address these challenges, we propose a novel training-free tri-branch denoising framework that effectively decouples structure from style via attention manipulation. Specifically, we propose two core mechanisms. Semantic Injection employs self attention maps to separate an object's spatial structure from its visual appearance. Style Guidance adapts advanced attention based style transfer techniques to the composition task for the first time. Comprehensive experimental results show that our method outperforms existing state-of-the-art approaches and achieves consistent improvements in structural consistency and stylistic coherence for image composition.

Prompt Yourself: Awakening Textual Semantics in 1D Visual Tokenizers