Qwen-Image-VAE-2.0 Technical Report

We present Qwen-Image-VAE-2.0, a suite of high-compression Variational Autoencoders (VAEs) that achieve significant advances in both reconstruction fidelity and diffusability. To address the reconstruction bottlenecks of high compression, we adopt an improved architecture featuring Global Skip Connections (GSC) and expanded latent channels. Moreover, we scale training to billions of images and incorporate a synthetic rendering engine to improve performance in text-rich scenarios. To tackle the convergence challenges of high-dimensional latent space, we implement an enhanced semantic alignment strategy to make the latent space highly amenable to diffusion modeling. To optimize computational efficiency, we leverage an asymmetric and attention-free encoder-decoder backbone to minimize encoding overhead. We present a comprehensive evaluation of Qwen-Image-VAE-2.0 on public reconstruction benchmarks. To evaluate performance in text-rich scenarios, we propose OmniDoc-TokenBench, a new benchmark comprising a diverse collection of real-world documents coupled with specialized OCR-based evaluation metrics. Qwen-Image-VAE-2.0 achieves state-of-the-art reconstruction performance, demonstrating exceptional capabilities in both general domains and text-rich scenarios at high compression ratio. Furthermore, downstream DiT experiments reveal our models possess superior diffusability, significantly accelerating convergence compared to existing high-compression baselines. These establish Qwen-Image-VAE-2.0 as a leading model with high compression, superior reconstruction, and exceptional diffusability.

Qwen-Image-2.0 Technical Report

We present Qwen-Image-2.0, an omni-capable image generation foundation model that unifies high-fidelity generation and precise image editing within a single framework. Despite recent progress, existing models still struggle with ultra-long text rendering, multilingual typography, high-resolution photorealism, robust instruction following, and efficient deployment, especially in text-rich and compositionally complex scenarios. Qwen-Image-2.0 addresses these challenges by coupling Qwen3-VL as the condition encoder with a Multimodal Diffusion Transformer for joint condition-target modeling, supported by large-scale data curation and a customized multi-stage training pipeline. This enables strong multimodal understanding while preserving flexible generation and editing capabilities. The model supports instructions of up to 1K tokens for generating text-rich content such as slides, posters, infographics, and comics, while significantly improving multilingual text fidelity and typography. It also enhances photorealistic generation with richer details, more realistic textures, and coherent lighting, and follows complex prompts more reliably across diverse styles. Extensive human evaluations show that Qwen-Image-2.0 substantially outperforms previous Qwen-Image models in both generation and editing, marking a step toward more general, reliable, and practical image generation foundation models.

KEVER^2: Knowledge-Enhanced Visual Emotion Reasoning and Retrieval

Understanding what emotions images evoke in their viewers is a foundational goal in human-centric visual computing. While recent advances in vision-language models (VLMs) have shown promise for visual emotion analysis (VEA), several key challenges remain unresolved. Emotional cues in images are often abstract, overlapping, and entangled, making them difficult to model and interpret. Moreover, VLMs struggle to align these complex visual patterns with emotional semantics due to limited supervision and sparse emotional grounding. Finally, existing approaches lack structured affective knowledge to resolve ambiguity and ensure consistent emotional reasoning across diverse visual domains. To address these limitations, we propose \textbf{K-EVER\textsuperscript{2}}, a knowledge-enhanced framework for emotion reasoning and retrieval. Our approach introduces a semantically structured formulation of visual emotion cues and integrates external affective knowledge through multimodal alignment. Without relying on handcrafted labels or direct emotion supervision, K-EVER\textsuperscript{2} achieves robust and interpretable emotion predictions across heterogeneous image types. We validate our framework on three representative benchmarks, Emotion6, EmoSet, and M-Disaster, covering social media imagery, human-centric scenes, and disaster contexts. K-EVER\textsuperscript{2} consistently outperforms strong CNN and VLM baselines, achieving up to a \textbf{19\% accuracy gain} for specific emotions and a \textbf{12.3\% average accuracy gain} across all emotion categories. Our results demonstrate a scalable and generalizable solution for advancing emotional understanding of visual content.

Balanced Token Pruning: Accelerating Vision Language Models Beyond Local Optimization

Large Vision-Language Models (LVLMs) have shown impressive performance across multi-modal tasks by encoding images into thousands of tokens. However, the large number of image tokens results in significant computational overhead, and the use of dynamic high-resolution inputs further increases this burden. Previous approaches have attempted to reduce the number of image tokens through token pruning, typically by selecting tokens based on attention scores or image token diversity. Through empirical studies, we observe that existing methods often overlook the joint impact of pruning on both the current layer's output (local) and the outputs of subsequent layers (global), leading to suboptimal pruning decisions. To address this challenge, we propose Balanced Token Pruning (BTP), a plug-and-play method for pruning vision tokens. Specifically, our method utilizes a small calibration set to divide the pruning process into multiple stages. In the early stages, our method emphasizes the impact of pruning on subsequent layers, whereas in the deeper stages, the focus shifts toward preserving the consistency of local outputs. Extensive experiments across various LVLMs demonstrate the broad effectiveness of our approach on multiple benchmarks. Our method achieves a 78% compression rate while preserving 96.7% of the original models' performance on average.