ECCV2026 Image Generation AI paper notes paper summaries Diffusion Models Super-Resolution Text-to-Image Image Editing Alignment/RLHF Reinforcement Learning

🎨 Image Generation¶

🎞️ ECCV2026 · 64 paper notes

📌 Same area in other venues: 📷 CVPR2026 (492) · 🔬 ICLR2026 (353) · 💬 ACL2026 (5) · 🧪 ICML2026 (141) · 🤖 AAAI2026 (79) · 🧠 NeurIPS2025 (221)

🔥 Top topics: Diffusion Models ×21 · Super-Resolution ×5 · Text-to-Image ×4 · Image Editing ×3 · Alignment/RLHF ×3

A Scalable Vector Graphics Latent Space: SLS compresses each SVG path, including its geometry and style, into a decodable dense vector, preserving strong reconstruction quality while reducing the average captioning input from 2432.02 XML tokens to 15.56 path embeddings; its main contribution is a vector-graphics representation interface rather than a new text-to-SVG generator.
AccelAes: Accelerating Diffusion Transformers for Training-Free Aesthetic-Enhanced Image Generation: AccelAes uses aesthetic prompt semantics and cross-attention to identify regions deserving precise updates, combines sparse computation with spatial CFG and step-level prediction reuse, and reduces Lumina-Next latency from 12.37 to 5.86 seconds while increasing ImageReward from 0.7518 to 0.8410 without training.
Accelerated Likelihood Maximization for Diffusion-based Versatile Content Generation: ALM proposes a completely training-free sampling strategy for diffusion models. It performs explicit likelihood maximization optimization on unobserved regions during the reverse diffusion process, and leverages the property that "updates of adjacent iterations are approximately equal" to collapse \(N\) iterations into a single update. It significantly outperforms prior state-of-the-art methods at zero training cost across multi-modal tasks including image inpainting/outpainting, human motion completion, 3D texture generation, and long video generation.
Accelerating Diffusion Transformers with Gaussian Process Rectified Feature Cache: GP-Refiner treats full-compute features on an accelerated trajectory as noisy observations, corrects cache predictions with an online Gaussian process, and triggers recomputation through posterior uncertainty; on Qwen-Image, it raises PSNR from 28.58 to 29.48 while reducing computation relative to a fixed-interval TaylorSeer configuration.
Achieving Subcategorical Erasure in Text-to-Image Models: SURE starts from a single parent-category target, maps that category and retrieved related concepts toward an empty prompt, preserves neutral predictions, and regularizes local latent variation, reaching a 10.00% subcategory detection rate on Guns without establishing complete erasure of every subcategory.
ActionParty: Multi-Subject Action Binding in Generative Video Games: ActionParty jointly predicts a shared game video and each character's coordinates, separating action ownership through attention masks and grounding subjects through spatial RoPE; across 46 Melting Pot games with up to 7 players, it raises movement accuracy from Text-Action's 0.158 to 0.779.
AdaBridge-SR: Adaptive Bridge Matching for Real-World Image Super-Resolution: AdaBridge-SR predicts where to inject noise and when to enter high-variance bridge states from the degraded input, using shared weights for deterministic one-step super-resolution and stochastic few-step refinement: on RealSR, it reaches 25.2152 dB PSNR in one step and 0.6847 MANIQA in two steps, while additional steps still incur a fidelity cost.
Adaptive Noise Covariance Scheduling under Riemannian Metrics for Diffusion Models: The paper separates blue-to-white noise evolution into a covariance path and a per-image transition speed, combining a Bures-Wasserstein geodesic with a latent-predicted temperature to improve image-generation FID and KID without replacing the BNDM denoising backbone.
Advancing WordArt-Oriented Scene Text Recognition: Datasets and Methods: This paper systematically advances WordArt-oriented scene text recognition (WATER) from both data and model perspectives. On the data side, a 2M synthetic WordArt dataset named WATER-S is constructed, consisting of a rendering-based subset WATER-T and a generative-model-based subset WATER-Z. On the model side, WATERec is proposed, supporting arbitrary-shape input and autoregressive decoding. It achieves over 90% accuracy (90.40%) on WordArt-Bench for the first time, significantly outperforming general VLMs and OCR-specific VLMs.
Adversarial Score Distillation for Stable One-Step Diffusion in Real-World Image Super-Resolution: ASDSR reorganizes the generator's adversarial score guidance, computes the discriminator penalty on ground-truth latents, and adds lightweight CLIP supervision for one-step 4x real-world super-resolution, reaching 28.77 PSNR, 0.7973 SSIM, and 0.2896 LPIPS on DrealSR without leading every perceptual metric.
Affordance-Guided Diffusion Prior for 3D Hand Reconstruction: AffHandGen conditions a hand-pose diffusion prior on detailed descriptions of how an object is grasped and used, together with image features, refining occluded joints while retaining visible 2D evidence and reducing WiLoR's high-occlusion PA-MPJPE on HOGraspNet from 12.41 to 9.37 mm.
AlignMorph: Tuning-Free Diffusion Image Morphing via Explicit Semantic Transport: AlignMorph aligns endpoint latents through semantic optimal transport before diffusion generation, reducing Morph4Data FID from 203.32 for reproduced FreeMorph to 188.44 without per-pair tuning, while increasing runtime from 90 to 115 seconds per image pair.
Allo{SR}2: Rectifying One-Step Super-Resolution to Stay Real via Allomorphic Generative Flows: AlloSR2 inserts low-resolution latents at an intermediate point of a pretrained generative flow, uses FATC to straighten the local restoration trajectory and ATM to align it with a parameter-shared generative trajectory, and performs 4x super-resolution with one velocity prediction, improving several realism metrics without targeting the highest pixel fidelity.
Analyzing and Improving Training-Free Fast Sampling of Text-to-Image Diffusion Models: TORS identifies the outer time schedule as the main bottleneck in few-step diffusion generation and places sampling times by equally partitioning the total rotation of projected three-dimensional trajectories, improving 10-step Flux.1-Dev Image Reward from 0.71 to 0.97 without matching the 50-step baseline on every metric.
Anchoring and Steering Diffusion: Enhancing the Faithfulness of Text-to-Image Generation at Inference Time: Without training model weights, AnchorSteer uses CLIP priors and a frozen diffusion model to optimize initialization, then diagnoses, removes, and completes semantic content during denoising, improving SDXL GenEval from 55.933 to 63.393 at a substantial inference cost.
Anchoring on Reality: Breaking the Pseudo-Target Ceiling in Makeup Transfer: Proposes the ART two-stage DiT framework, which breaks the pseudo-target ceiling through a differentiable reality anchoring loop—reconstructing the real reference from a bare-skin reference to allow weight gradients to penalize omissions—achieving high-fidelity transfer of complex makeup such as dense stickers and face painting at 2K resolution.
AnyStyle: A Single LoRA is Sufficient for Image-Guided Style Transfer: AnyStyle learns appearance with one LoRA per reference style and injects content structure through timestep-matched query modulation, achieving CLIP-T of 32.5573 and CLIP-Style of 67.9291 across 400 content-style pairs while replacing the second, content-specific LoRA with training-free guidance, though it does not lead every content-preservation metric.
ARGENT: Adaptive Hierarchical Image-Text Representations: ARGENT replaces unstable cone-aperture constraints in a HyCoCLIP hyperbolic dual encoder with adaptive angular entailment and norm regularization, while redesigning hierarchical evaluation through PEP; its Large model improves ImageNet accuracy by 1.7 percentage points and PEP AUC by 1.5 points over HyCoCLIP-L, although gains do not hold for every scale and metric.
ARMS: Anchor–Relational Motion Streaming for Seamless Solo-Social Motion Transitions: ARMS continuously extends one person's trajectory while locating a partner relative to that person, using causal segment-wise denoising and relational gating to enter or leave interactions without restarting generation; solo-interaction transition PJ falls from 3.131 for adapted InterMask to 0.077, although ARMS does not lead every generation-quality metric.
Attribute Token Arithmetic: Disentangled and Continuous Semantic Control for Visual Autoregressive Models: ATA learns and subtracts two reconstruction directions from one reference image and prompts differing only in a target attribute, enabling compositional strength control in frozen Infinity text-conditioning space; in a same-backbone comparison, mean ΔVQA rises from 0.532 with the Concept Slider strategy to 0.567, and mean I-LPIPS from 0.879 to 0.895.
Autoregressive Image Generation Needs Only a Few Lines of Cached Tokens: After each generated image line, LineAR uses that line's queries to identify historical visual tokens unlikely to matter for the next line, while protecting initial anchors and recent lines; at a 1/6 visual cache budget, it improves LlamaGen-XL ImageNet FID from 2.77 to 2.68 and Janus-Pro-1B COCO FID from 23.85 to 22.86 without retraining.
AVSR-Diff: Scale-Agnostic Diffusion Priors for Temporally Consistent Arbitrary-Scale Video Super-Resolution: This paper proposes AVSR-Diff, which decouples the scale-agnostic latent denoising of diffusion models from continuous coordinate decoding. It utilizes a Temporal Gated Feature Recurrence (TGFR) module for cross-frame feature alignment to suppress flickering, and a Scale-Aware Fourier Refinement (SAFR) module for scale-adaptive frequency modulation in the Fourier domain. This architecture enables the simultaneous achievement of high generation quality and stable temporal consistency for arbitrary-scale VSR, while maintaining a computational overhead that does not scale with the output resolution.
A²-Edit: Precise Reference-Guided Image Editing of Arbitrary Objects and Ambiguous Masks: A²-Edit augments FLUX.1 Fill with joint attention-FFN LoRA experts, a training curriculum that progressively relaxes segmentation masks into bounding boxes, and 500,104 cross-category training samples to retain reference identity and structure under coarse masks, improving UniEdit rough-mask DINO-I from Insert Anything's 56.13 to 62.28.
Be Tangential to Manifold: Discovering Riemannian Metric for Diffusion Models: The paper turns a pretrained diffusion model's score Jacobian into a training-free Riemannian metric that favors motion along the data manifold during interpolation and guidance, reducing Animal Faces-HQ interpolation FID from GeoDiff's 25.80 to 21.01 and generation FID from 17.28 to 16.04 at CFG scale 12.5.
Beyond Absolute Scores: Relative Edit-induced Difference for Generalizable Image Aesthetic Assessment: Instead of fitting absolute scores for isolated images, RED-Aes learns causal aesthetic changes from controllably edited pairs and reaches 0.7442/0.7322 average PLCC/SRCC across five zero-shot cross-domain benchmarks, outperforming both specialist and generalist models.
C3-Bench: A Context-Aware Change Captioning Benchmark: C3-Bench is a comprehensive benchmark for the context-aware change captioning task. It features 4,996 human-annotated image pairs spanning 51 real-world change contexts across 4 major domains. It introduces the LLM-as-Judge framework and a reversibility metric to change captioning evaluation for the first time. Benchmarking 32 models reveals critical blind spots, such as the complete failure of traditional models in open-world scenarios and systematic perceptual, spatial, and positional biases in large multimodal models (LMMs).
Continuous Speculative Decoding for Autoregressive Image Generation: This paper is the first to extend speculative decoding from discrete to continuous distributions. It enhances the draft-target acceptance rate through denoising trajectory alignment, and resolves the sampling bottleneck of the correction distribution (which lacks an analytical expression) via acceptance-rejection sampling. It achieves over 2x wall-clock speedup on multiple continuous visual AR models including MAR, xAR, and Harmon while maintaining generation quality, completely training-free.
Curvature-Adaptive Consistency Flow Matching: Autonomous Trajectory Optimization via Reinforcement Learning: CACFM reformulates consistency distillation as a dynamic decision-making process. It utilizes a lightweight RL agent to automatically detect and prioritize high-curvature bottleneck segments along the PF-ODE trajectory for training. Combining flow-adapted DMD and adversarial consistency loss, it achieves state-of-the-art results on FLUX and SDXL with 4-step inference, delivering an FID over 2 points lower than FLUX-schnell.
Diffusion Integrated Gradients: Controllable Path Generation for Flexible Feature Attribution: DiffIG reformulates attribution path generation in Integrated Gradients (IG) as a conditional generative modeling problem. It uses a diffusion model to learn the path distribution generated by the Stick-Breaking Process, and performs dual-guided sampling based on faithfulness and complexity during inference to generate adaptive non-linear integration paths. This approach significantly outperforms existing path attribution methods on the DiffID metric across Oxford-IIIT Pet and Mini-ImageNet.
Divide and Conquer: Decoupled Representation Alignment for Multimodal World Models: M2-REPA extends single-modality REPA to multimodal video generation for the first time by decoupling intermediate features of diffusion models into modality-specific subspaces to perform representation alignment with modality experts (e.g., DINOv2, DepthAnythingV2, SAM2) individually, while using CKA regularization to suppress cross-modality redundancy, achieving significant improvements in tri-modal (RGB-depth-mask) generation.
DivRL: Disentangled Self-Similarity Rewards for Diverse Subject-Driven Generation: DivRL proposes a Flow-GRPO-based post-training framework that measures structural diversity and identity consistency using disentangled visual features, employing a two-stage "explore-and-suppress" optimization strategy to enable the model to simultaneously generate images with high identity preservation and high structural diversity.
Don't Settle at the Mode! Mitigating Diversity Collapse in Pretrained Flow Models via Feature Self-Guidance: This paper discovers that pretrained flow matching models (e.g., FLUX) experience internal MMDiT feature collapse when generating multiple samples from the same conditional input, which leads to output diversity collapse. It proposes a training-free, inference-time Feature Self-Guidance mechanism. By iteratively dispersing the intermediate features of samples within a batch and subsequently re-projecting them back onto the manifold through the same MMDiT block, this approach significantly enhances output diversity with minimal inference overhead while maintaining prompt alignment.
DriftScope: Measuring The Hidden Effects of Diffusion Model Adaptation: This paper demonstrates that text-to-image diffusion models undergoing adaptation (both concept customization and concept erasing) systematically damage other semantically unrelated concepts (with zero-shot classification accuracy of the worst-performing category dropping by up to 18.9 percentage points), while aggregate metrics such as FID/KID are structurally blind to this degradation. The authors propose DriftScope, a soft-prompt-optimization-based token-level concept drift diagnostic tool that ranks the most severely affected tokens by maximizing the divergence of cross-attention maps between two checkpoints.
DTI: Dynamic Trajectory Initialization for Generative Face Video Super-Resolution: DTI proposes a "Dynamic Trajectory Initialization" paradigm, redefining generative face video super-resolution from "full generation" to "input-driven targeted restoration." By coupling a Discriminative Guider (DG) with SNR alignment theory, it dynamically determines the starting point of diffusion sampling for each low-quality input. With only minor fine-tuning, this approach comprehensively outperforms existing methods in fidelity, perceptual quality, and inference efficiency.
Dual-End Consistency Model: DE-CM achieves SOTA performance on ImageNet 256 with a 1-step generation FID of 1.70. It optimizes three key sub-trajectories (consistency trajectory, instantaneous velocity trajectory, and noise-to-noise trajectory) selected from the entire PF-ODE path, stabilizes training with flow matching as boundary regularization, and mitigates error accumulation via an N2N mapping.
Edit in 2D, Verify in 3D: Reinforcement Learning for Multi-view Consistent Scene Editing: RL3DEdit leverages the GRPO reinforcement learning framework, utilizing confidence maps and pose predictions from the VGGT 3D foundation model as geometric-aware reward signals. Without requiring paired 3D editing data, it anchors the 2D editor FLUX-Kontext onto a 3D consistent manifold, achieving highly efficient multi-view consistent 3D scene editing in a single inference pass.
FedOT: Ownership Verification and Leakage Tracing via Watermarks for Federated LDMs: FedOT proposes to embed block-based watermarks coupled with Latent Vector Transformation (LVT) during the training of federated latent diffusion models. This marks the first framework to simultaneously achieve ownership verification and malicious client tracing, ensuring that even if an attacker replaces the watermarked VAE, the watermark cannot be removed without severely degrading the generation quality.
Few-Shot Synthetic Image Attribution: Identifying Unseen Generators with Limited Samples: This paper defines a brand-new task of "few-shot synthetic image attribution"—identifying unseen image generators with only 10 reference images per class. It introduces the OmniFake dataset containing 45 independent generators (excluding PEFT variants) and a dual-path contrastive learning baseline OmniDFA, which significantly outperforms prior methods in both attribution and detection.
Focusing on What Matters: Saliency-Harnessing Accurate Routing for Diffusion MoE: SharpMoE proposes a plug-and-play post-training framework that uses the clean latent variable \(x̂₀\) predicted from the previous denoising step instead of the noisy latent variable as the router input. This addresses the core issue in diffusion MoEs where "noisy routing" leads to a failure in identifying salient tokens. Augmented with a trajectory routing loss to globally align cumulative computation allocation, it significantly improves the generation quality of various converged pre-trained diffusion MoE models with only 100K post-training steps.
FUMO: Prior-Modulated Diffusion for Single Image Reflection Removal: FUMO proposes a prior-modulated diffusion framework for single image reflection removal. It extracts VLM-driven reflection intensity priors and multi-scale high-frequency priors from the mixed image to spatially and adaptively modulate the conditional residual injection of ControlNet via a gated mechanism. Guided by a coarse-to-fine two-stage process (one-step denoising diffusion coarse restoration + a fine-grained refinement module), it achieves competitive quantitative results and consistent perceptual quality improvements on three standard benchmarks and real-world scenes.
Gen2Balance: Generative Balancing for Long-Tailed Video Action Recognition: Gen2Balance utilizes a text-to-video generative model (WAN 2.1) to synthesize tail-class samples for long-tailed video action recognition. It ensures the diversity and semantic correctness of generated videos through a VLM-driven (Gemini 2.5 Pro) three-tier prompting pipeline (diversity expansion + Action Profile disambiguation + In-Context Exemplars anchoring). Combined with a two-stage training strategy (learning representations on mixed synthetic/real data, and calibrating the classifier using only real data with real-frequency priors), Gen2Balance outperforms the strongest long-tailed baselines by 7.0% and 5.1% on K100-LT and UCF-LT respectively, and achieves a 31.9% improvement on RareAct rare actions.
GeoEdit: Geometry-Aware Object Editing via Dual-Branch Denoising: GeoEdit proposes a training-free "lift-manipulate-render-denoise" pipeline. It first reconstructs a single-view image into 3D in a decoupled manner, renders a geometrically aligned coarse proxy image according to user-specified 3D transformations (translation, rotation, scaling), and then employs "dual-branch denoising" to inject the 3D constraints solely into the foreground in a variance-homogeneous manner within a denoising window. This allows the background to generate freely, simultaneously adhering to physical constraints like perspective and occlusion while balancing foreground rigidity and background realism.
In-context Region-based Drag: Drag Any Region to Any Shape: ICRDrag reformulates region-based drag as an in-context learning task. By taking the source image, source region mask, and target region mask as a unified conditional input to the DiT, the target image generation is guided via two novel regularization losses: Image-Mask Attention Consistency (IMAC) and Source-Target Attention Correspondence (STAC). Additionally, the paired 287k-sample Paired Region Dataset (PRD) is constructed. It significantly outperforms existing methods in editing accuracy, visual fidelity, and user preference.
Intermediate Text Representation Guided Text-to-Image Generation for Enhancing One-and-Only Alignment: This paper demonstrates through information-theoretic analysis that the intermediate layers of text encoders retain more conceptual mutual information than the final embeddings. It proposes IR-guided diffusion, which injects intermediate hidden states into the conditioning signals during the early stages of denoising to recover attributes suppressed by strong priors. Requiring no training or external models, it improves VQAScore by up to 19.1 percentage points on OAO-AttackBench while maintaining generation quality and human preference.
ISAC: Training-Free Instance-to-Semantic Attention Control for Multi-Instance Generation: ISAC proposes a hierarchical, instance-first, training-free attention control method. It first leverages self-attention to stabilize instance layouts in early denoising steps, and then binds additional semantics within the established instance regions, significantly improving count accuracy and semantic separation in multi-instance generation with diffusion models.
Learn Once, Edit Anywhere: Visual Direction Transfer for Diffusion Models: ViDiT proposes a "learn once, edit anywhere" framework: it optimizes a continuous latent direction d from a small number of image edit pairs (before/after) and injects this direction into the CFG inference process of a diffusion model. This achieves zero-shot, disentangled, and intensity-controllable fine-grained attribute editing on arbitrary images, without fine-tuning the base model or requiring image-specific optimization.
MEPA: Multi-Scale Representation Alignment for Visual Autoregressive Modeling with Mixture of Experts: To address the issues of multi-scale representation learning conflicts and early-scale semantic error propagation caused by a shared architecture in Visual AutoRegressive (VAR) models, MEPA proposes a Scale-aware Token-routed Mixture of Experts (STMoE) to decouple model capacity across different scales. Furthermore, it incorporates self-supervised visual features (DINOv3) to perform semantic alignment on the accumulated residual aggregated representations of early scales. On ImageNet 256x256, MEPA outperforms the VAR baseline using only half the training epochs, reducing FID by 0.63–0.69.
MGI: Member vs Generated Inference: This paper proposes a new task, Member vs Generated Inference (MGI): given a sample and a generative model, determine whether the sample is a training member of the model or an output generated by the model itself. The authors design a three-stage cascade method, DCB, which filters generated samples using autoencoder reconstruction/quantization errors, performs membership inference using the latent space generator, and finally traces the source by comparing conditional probabilities across generators. DCB significantly outperforms existing MIA and attribution methods on both Image Autoregressive (IAR) and diffusion models.
MIMFlow: Integrating Masked Image Modeling with Normalizing Flows for End-to-End Image Generation: MIMFlow proposes an end-to-end framework that extracts semantic latent representations using a learnable token bottleneck on masked images in a VAE encoder. This allows the normalizing flow to focus on modeling low-frequency semantic manifolds, while the decoder handles high-frequency texture synthesis, achieving an FID of 2.50 and a linear probing accuracy of 71.3% on ImageNet 256x256 using only 128 tokens (a 50% reduction compared to standard models).
MirrorPPR: Exemplar-Based Portrait Photo Retouching: MirrorPPR proposes an exemplar-guided portrait structural retouching task. Through a dedicated editing operation extractor (MAE + R-Former), a connector, a DiT dual-stream backbone (Qwen-Image-Edit-2511), and a LoRA adaptation pipeline, combined with a 47-million-pair dataset generated by a self-augmentation data strategy, it achieves precise cross-identity transfer of fine structural retouching operations for the first time.
Not All Prediction Targets Keep Training-Free Diffusion Guidance on the Manifold: This work theoretically proves that the prediction targets of diffusion models (\(\epsilon\)-prediction / \(v\)-prediction / \(x\)-prediction) form a strict error amplification hierarchy under training-free guidance (TFG). Specifically, \(\epsilon\)-prediction causes samples to drift off the data manifold in high-noise steps due to the divergence of the \(1/t\) factor in its reconstruction formula, whereas \(x\)-prediction directly outputs clean images without error amplification. These hierarchical effects are verified across three dimensions: a controlled crossed-lines ablation study, a 143-class fine-grained bird benchmark, and style transfer. Additionally, Child FID is introduced as a manifold-aware evaluation metric.
Obliviate: Erasing Concepts from Autoregressive Image Generation Models: This paper adapts the mature "negative guidance concept erasure" from diffusion models to autoregressive image generation models. By aligning conditional and pseudo-unconditional predictions via shared visual prefixes and applying KL distribution supervision over the entire token trajectory, it thoroughly erases concepts such as nudity, gore, and brand logos while causing minimal damage to model utility (e.g., reducing the nudity detection rate of RAB on Liquid from 91.58% to 3.15%).
OTCache: Optimal Transport for Geometry-Aware Caching in Diffusion Models: OTCache proposes a training-free acceleration framework for diffusion models. Based on optimal transport theory, it models cache scheduling under different inference budgets as continuous evolutionary trajectories in the policy space. Using a high-budget graph search reference schedule and a low-budget black-box search anchor as two endpoints, it predicts scheduling plans for any target budget through quantile interpolation in Wasserstein space. It achieves 4.5x, 4.7x, and 3.66x speedups on FLUX.1, Qwen-Image, and HunyuanVideo, respectively, consistently outperforming existing graph-search cache baselines.
Phase-Aligned RoPE for Mixed-Resolution Diffusion Transformer: This paper reveals for the first time the root cause of RoPE's failure in mixed-resolution DiTs: namely, attention scores are a sinusoidal periodic function of the relative token distance \(\kappa(\Delta)\). When tokens of different resolutions are mapped to a unified coordinate space, they land on different phases of the \(\kappa(\Delta)\) curve, causing systematic attention score distortion. Based on this, a training-free Phase-Aligned Mixed-Resolution Attention (PMA) is proposed. PMA calculates position offsets for each query-key pair at the original scale of the query, recovering a consistent phase reference. Combined with a lightweight Boundary Enhancement and Refinement module (BER), PMA achieves generation quality comparable to or even better than full resolution with 4x acceleration on the Wan video model and the FLUX image model.
Prompt2Effect: Training-Free Image-to-Video Model Specialization via LoRA Generation: Prompt2Effect proposes a weight-driven hypernetwork that explicitly utilizes frozen base model weights as conditioning and normalizes the LoRA parameter prediction space into an SVD energy-ordered format. It directly synthesizes effect-specific LoRA weights in a single forward pass, achieving zero-shot training-free effect control for I2V diffusion models. Inference requires only 3.3 seconds while achieving quality comparable to full LoRA fine-tuning.
Rethinking Garment Conditioning in Diffusion-based Virtual Try-On: Decouple, Don't Denoise: DeCo-VTON reveals for the first time, through visualization of the noise prediction behavior of dual-UNet reference networks, the root cause behind the failure of full-parameter fine-tuning in spatial-concatenation-based virtual try-on: three functional conflicts caused by the coupling of garment condition and the denoising process. Correspondingly, three decoupling design principles are proposed (garment-free guidance, decoupled loss, and clean latent anchoring). Without modifying the network architecture, DeCo-VTON achieves or surpasses the try-on quality of the 1.8B-parameter state-of-the-art dual-UNet model Leffa, using only a single 860M-parameter UNet.
Semantic Browsing: Controllable Diversity for Image Generation: This paper proposes the Semantic Browsing framework, which utilizes a multi-agent VLM pipeline to automatically expand user prompts into structured JSON scene representations. It then systematically mutates semantic dimensions to generate hierarchical diverse image galleries, achieving navigable, interpretable, and controllable diversity generation.
SONIC: Spectral Optimization of Noise for Inpainting with Consistency: SONIC proposes a training-free diffusion-based image inpainting method. The core idea is to optimize the initial noise map so that the reconstructed image after denoising faithfully restores unmasked regions. It bypasses expensive backpropagation via denoising trajectory linearization and uses Adam in the spectral domain for preconditioned optimization to achieve stable convergence, outperforming existing methods (including the specially trained BrushNet) across three standard inpainting datasets.
Spanning the Visual Analogy Space with a Weight Basis of LoRAs: LoRWeB replaces a single LoRA with a set of learnable LoRA bases for visual analogy editing: a lightweight CLIP encoder dynamically calculates attention weights based on the input analogy pair, and combines \(N\) basis LoRAs into a task-specific mixed LoRA injected into Flux.1-Kontext. This achieves a significant generalization improvement on unseen transformation types while increasing inference time by only 3% compared to a single LoRA.
SRUM: Fine-Grained Self-Rewarding for Unified Multimodal Models: SRUM proposes a self-rewarding post-training framework where the comprehension module of a Unified Multimodal Model (UMM) acts as an internal evaluator, providing fine-grained corrective signals to the generation module through a dual global-local reward system, thereby significantly improving the quality of complex text-to-image generation.
Symbiotic-MoE: Unlocking the Synergy between Generation and Understanding: Symbiotic-MoE resolves the gradient conflict between generation and understanding tasks within a native MoE Transformer architecture through modality-aware expert decoupling and a progressive training strategy. With zero parameter overhead, it achieves synergistic improvement in both capabilities—yielding generation quality superior to MoT/Bagel, while understanding capabilities (MMLU +20.1%, OCRBench +13.8%) even surpass the understanding-only baseline.
Text-Conditioned Background Generation for Editable Multi-Layer Documents: By treating a multi-page document as a separable "text / image / background" three-layer structure, this work proposes a training-free framework to regenerate only the background layer. By applying soft attenuation to the foreground regions in the diffusion latent space, the text is protected. A contrast-driven algorithm automatically adds semi-transparent backing cards to the text to guarantee readability, while a recursive summarization-instruction memory maintains thematic style across the entire document. Ultimately, this approach significantly outperforms BAGEL and GPT-5 in readability (WCAG 99.75% / OCR 0.97) and user preference (83.57%).
Universal Image Immunization against Diffusion-based Image Editing via Semantic Injection: To defend against malicious editing of diffusion models, this paper trains an image-agnostic universal adversarial perturbation (UAP). Through dual losses of "injecting target semantics + suppressing source semantics", the diffusion model is forced to misidentify the protected image as a pre-defined target concept (e.g., misinterpreting a dog as "Ronaldo"). Consequently, during inference, a simple addition operation is sufficient to block unauthorized editing, incurring near-zero computational overhead.
VTEdit-Bench: A Comprehensive Benchmark for Multi-Reference Image Editing Models in Virtual Try-On: VTEdit-Bench presents the first benchmark to systematically evaluate the performance of general-purpose multi-reference image editing models in virtual try-on (VTON) scenarios. It includes 24,220 evaluation pairs across 5 progressively complex tasks, and features VTEdit-QA (a reference-aware evaluation agent based on GPT-4o) for fine-grained automatic assessment. Experimental results demonstrate that general image editors match the performance of dedicated models in classic try-on tasks and exhibit stabler generalization in complex scenarios, though a significant gap remains in multi-garment composition control.