A safety direction estimated in a source LLM is transported to a target generator through lightweight alignment on benign data alone, matching native safety performance without any target-side unsafe data.
super hub Mixed citations
Qwen-Image Technical Report
Mixed citation behavior. Most common role is background (46%).
311
Pith papers citing it
Background
46% of classified citations
abstract
We present Qwen-Image, an image generation foundation model in the Qwen series that achieves significant advances in complex text rendering and precise image editing. To address the challenges of complex text rendering, we design a comprehensive data pipeline that includes large-scale data collection, filtering, annotation, synthesis, and balancing. Moreover, we adopt a progressive training strategy that starts with non-text-to-text rendering, evolves from simple to complex textual inputs, and gradually scales up to paragraph-level descriptions. This curriculum learning approach substantially enhances the model's native text rendering capabilities. As a result, Qwen-Image not only performs exceptionally well in alphabetic languages such as English, but also achieves remarkable progress on more challenging logographic languages like Chinese. To enhance image editing consistency, we introduce an improved multi-task training paradigm that incorporates not only traditional text-to-image (T2I) and text-image-to-image (TI2I) tasks but also image-to-image (I2I) reconstruction, effectively aligning the latent representations between Qwen2.5-VL and MMDiT. Furthermore, we separately feed the original image into Qwen2.5-VL and the VAE encoder to obtain semantic and reconstructive representations, respectively. This dual-encoding mechanism enables the editing module to strike a balance between preserving semantic consistency and maintaining visual fidelity. Qwen-Image achieves state-of-the-art performance, demonstrating its strong capabilities in both image generation and editing across multiple benchmarks.
hub tools
citation-role summary
background 34
baseline 23
method 10
dataset 2
citation-polarity summary
claims ledger
- abstract We present Qwen-Image, an image generation foundation model in the Qwen series that achieves significant advances in complex text rendering and precise image editing. To address the challenges of complex text rendering, we design a comprehensive data pipeline that includes large-scale data collection, filtering, annotation, synthesis, and balancing. Moreover, we adopt a progressive training strategy that starts with non-text-to-text rendering, evolves from simple to complex textual inputs, and gradually scales up to paragraph-level descriptions. This curriculum learning approach substantially
authors
co-cited works
representative citing papers
The work creates the first dataset and baseline for generating emission textures on 3D objects to reproduce glowing materials from input images.
OP-GRPO is the first off-policy GRPO method for flow-matching models that reuses trajectories via replay buffer and importance sampling corrections, matching on-policy performance with 34.2% of the training steps.
Cornfigurator is the first automated deployment planner for generic any-to-any multimodal models that explores the full range of colocation-to-disaggregation strategies and delivers 1.12x to 6.32x higher goodput than existing systems or expert plans.
OrbitQuant is a data-agnostic PTQ technique for DiTs that uses RPBH rotation in a normalized basis to enable a single codebook across all inputs, achieving SOTA low-bit performance on FLUX.1, CogVideoX and similar models.
Introduces VICIS task and training framework for inferring visual concepts from image sets, with experiments showing better accuracy, diversity, and generalization than standard VLMs on synthetic and ImageNet data.
A new pipeline using canonical LoRAs for view synthesis, deformable 3D Gaussian splatting anchored on D-SMAL, and generative repair to produce animatable 3D dogs from single wild images without 3D supervision.
Introduces VidPair-Halluc benchmark of 1K background-controlled adversarial video pairs and 11K QA pairs generated via PairFlow pipeline to evaluate hallucination in LVMs.
PhyEditBench is a new benchmark for physics-aware image editing with real and synthetic instances plus a training-free PhyWorld baseline that uses test-time scaling to outperform SOTA models.
InterleaveThinker is the first multi-agent pipeline enabling interleaved generation in any image generator through planner-critic agents, SFT on custom datasets, and GRPO RL with accuracy and step-wise rewards.
CineOrchestra unifies control of subjects, events, cameras, and shot transitions in cinematic video generation through entity-centric conditioning primitives and parameter-free coordinated rotary embeddings.
The paper constructs the VIBE benchmark and evaluates six visual in-context learning models on 14 datasets, 12 tasks, and 106 combinations under a unified one-shot protocol, revealing limitations and failure modes.
ImageTime is a benchmark that probes image generation models' visual world modeling by requiring coherent four-state sequences in single images, scored via VLM judge.
SSR-Merge merges LoRAs via subspace construction, inverse correlation decorrelation, and directional steering, shown to match the OLS solution with a streaming implementation that outperforms prior merging methods.
PoisonLoRA demonstrates ~100% attack success rates for stealthy LoRA poisoning via concept hijacking and task injection on real platforms, with robustness to base model transfer and multiple remixes.
ZIPP conditions diffusion models on LLM-rewritten prompts derived from graph-mined natural-language personas to achieve zero-shot personalization, reporting 13-20% gains and 79% human preference win rate over generic outputs.
CBS partitions the diffusion timeline into segments of equal approximation burden via Dirichlet energy and trajectory acceleration monitors estimated by an auxiliary model, yielding higher synthesis quality at fixed per-step cost across SiT, JiT and UNet backbones.
RED-Aes learns aesthetic changes from edit-induced image pairs and a new RED-20k dataset via three-stage relative ranking training, claiming SOTA generalization over absolute MOS regression.
GeM-NR performs multi-view consistent nonrigid editing by aligning depth-derived point clouds between edited and unedited scenes then refining projections conditioned on the original query view.
UniCAD supplies a unified multi-modal benchmark and an end-to-end MLLM that performs reconstruction, generation, and QA on CAD data, reporting SOTA results on UniCAD and Fusion360.
MetaPoint represents 2D coordinates as special tokens in visual generative models to enable precise spatial control using existing positional encodings without architectural modifications.
Introduces the Impostor benchmark dataset for localizing AIGC image manipulations via agent curation and the PANet model that uses phase and semantic consistency for better detection.
A large-scale empirical study across tokenizers and diffusion backbones identifies Velocity Irreducible Variance (VIV) as one of the most stable predictors of latent diffusion generation quality.
The SIU²A framework evaluates scientific images for error detection, repair feasibility, and correction quality, showing current multimodal systems have major limitations in preserving scientific validity.