🎨 Image Generation

🎞️ ECCV2026 · 328 paper notes

📌 Same area in other venues: 🧠 NeurIPS2026 (1) · 📷 CVPR2026 (490) · 🔬 ICLR2026 (352) · 💬 ACL2026 (5) · 🧪 ICML2026 (141) · 🤖 AAAI2026 (79)

🔥 Top topics: Diffusion Models ×94 · Text-to-Image ×25 · Alignment/RLHF ×23 · Image Editing ×20 · Multimodal/VLM ×12

MixGRPO: Unlocking Flow-based GRPO Efficiency with Mixed ODE-SDE

MixGRPO concentrates stochastic exploration and GRPO updates in a short window that moves from coarse to fine denoising stages, using deterministic sampling elsewhere; on FLUX.1-dev with HPDv2 evaluation, it improves preference scores with fewer optimized steps, while Table 1 reports approximately 49% and 71% lower training iteration time for the base version and the fastest Flash* version, respectively.

TIIF-Bench: How Does Your T2I Model Follow Your Instructions?

TIIF-Bench combines hierarchical short/long prompt pairs, attribute-level binary checks, and specialized text/style metrics, adapting Qwen3VL-8B into an interpretable evaluator that achieves 91.07% accuracy on a human-annotated validation set.

Distribution Matching Distillation Meets Reinforcement Learning

DMDR couples few-step diffusion distillation and preference optimization in a two-stage training process, allowing 4-step students to outperform multi-step teachers on several overall metrics, although preference gains still reduce generation diversity.

Towards Scalable Pre-training of Visual Tokenizers for Generation

VTP trains a visual tokenizer for reconstruction, visual semantics, and image-text alignment together, allowing additional tokenizer pretraining to improve downstream diffusion generation and achieving guided gFID 1.11 on ImageNet 256×256 with a 675M generator trained for 600 epochs.

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.

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.

GenAgent: Scaling Text-to-Image Generation via Agentic Multimodal Reasoning

GenAgent treats the generator as a frozen tool and trains one multimodal model through cold-start training and agentic reinforcement learning to rewrite prompts, inspect images, and reflect, raising FLUX's GenEval++ score from 0.325 to 0.561 with at most two generation rounds in the main experiments.

Query-Kontext: An Unified Multimodal Model for Image Generation and Editing

Query-Kontext transfers semantic and coarse visual planning from a VLM to a diffusion model through trainable multimodal query tokens, adds low-level image conditioning for detail, and unifies generation and editing with 0.88 on GenEval using rewritten prompts and 7.66 / 7.65 on the full English / Chinese GEdit-Bench splits.

RealGen: Photorealistic Text-to-Image Generation via Detector-Guided Rewards

RealGen turns synthetic-image detectors into realism rewards, first optimizing Qwen3 prompt rewriting and then FLUX image generation, achieving a 50.15% model-judged win rate against real photographs on RealBench with the full system, which does not establish indistinguishability for all human observers.

Learning on the Manifold: Unlocking Standard Diffusion Transformers with Representation Encoders

RJF replaces straight-line flow matching in representation latent spaces with spherical geodesic flow matching and weights training by curvature-dependent error propagation, enabling a standard-width, 131M-parameter DiT-B to achieve FID 4.95 without guidance and 3.37 with guidance on ImageNet 256 class-conditional generation, both after 200 epochs.

BlenderFusion: 3D-Grounded Visual Editing and Generative Compositing

BlenderFusion lifts image objects into editable geometry, expresses precise editing intent in Blender, and completes and fuses the scene with a dual-stream diffusion compositor, outperforming adapted Object 3DIT and Neural Assets on three video datasets while supporting fine-grained compositional edits beyond its training operations.

SAEdit: Token-Level Control for Continuous Image Editing via Sparse Autoencoder

SAEdit maps T5 token representations into a sparse semantic space, selects and aggregates scalable attribute directions, and schedules their injection to control image editing strength; its overall preference win rate against ConceptSlider (Flux) is 71% in a Flux user study, although disentanglement remains constrained by the generator's priors.

SD3.5-Flash: Distribution-Guided Distillation of Generative Flows

SD3.5-Flash places distribution matching on the student's own generation trajectory and combines split-timestep fine-tuning with pipeline optimization to distill SD3.5M into a four-step text-to-image model, improving selected preference and alignment metrics while supporting phones, without improving every quality metric.

Discrete Noise Inversion for Next-scale Autoregressive Text-based Image Editing

VARIN constructs inverse noise with a controllable preservation bias for next-scale visual autoregressive models, enabling editing without training or test-time optimization and reducing the HART Regeneration baseline's structure distance from 25.56 to 11.54 on PIE-Bench while improving background preservation and text alignment.

Exposure Bias Can Alleviate Itself via Directional and Frequency Rectification in Flow Matching

DEFAR simulates single-step inference during flow matching training and turns the model's own drift into directional rectification and frequency compensation signals, reducing SiT-B/4's ImageNet-256 FID with CFG from 31.64 to 29.81 at a comparable training time without changing the inference model architecture.

LightFusion: A Light-weighted, Double Fusion Framework for Unified Multimodal Understanding and Generation

LightFusion preserves the Qwen2.5-VL-7B understanding branch and Wan2.2-TI2V-5B generation branch while inserting layer-wise multimodal attention and combining source-image ViT and VAE representations, achieving 0.91 on GenEval with rewritten prompts and 3.77 on ImgEdit-Bench after approximately 35B training tokens.

Scientific Image Synthesis: Benchmarking, Methodologies, and Downstream Utility

SciGenBench measures whether scientific images preserve usable reasoning information, comparing direct pixel generation with an understand-plan-code ImgCoder workflow; Gemini-3-Pro-ImgCoder achieves a 77.87% image-level inverse validation rate under the default evaluator, and synthetic data also improves downstream multimodal reasoning.

Target-aware Image Editing via Cycle-consistent Constraints

FlowCycle makes the noise defining an editing starting point a per-image optimization variable and preserves unrelated content through dual consistency constraints over a source-to-target-to-source cycle, achieving lower Structure Distance and higher CLIP Score than FlowAlign on PIE-Bench at substantially greater editing time.

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.

Diffusion-SDPO: Safeguarded Direct Preference Optimization for Diffusion Models

Diffusion-SDPO adaptively scales the loser's backward gradient using its relationship with the winner gradient, addressing the failure of diffusion DPO in which the preference margin improves while both branches deteriorate; in the paper's FLUX.1-dev setting, GenEval increases from the base model's 0.66 to 0.71, whereas ordinary DPO at the same learning rate drops to 0.23.

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.

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.

SSDD: Single-Step Diffusion Decoder for Efficient Image Tokenization

SSDD trains a convolutional-transformer pixel-space decoder without a GAN and distills its multi-step teacher into a single-step student, allowing its small model to reduce KL-VAE's rFID from 0.87 to 0.46 while increasing decoding throughput from 705 to 1027 images/s on ImageNet at 256 x 256 with f8c4 latents.

CoCo: Code as CoT for Text-to-Image Preview and Rare Concept Generation

CoCo trains Bagel to generate and execute code into a structural draft before draft-guided image refinement, raising StructT2IBench overall accuracy from 4.69% to 73.52%, although the evidence primarily concerns structured visuals and text rendering rather than general rare-concept generation.

Continuous Adversarial Flow Models

CAFM compares real and predicted velocities in the discriminator's directional-derivative space to adversarially post-train continuous-time flow models, reducing guidance-free ImageNet 256px FID from 8.26 / 7.17 to 3.63 / 3.57 for SiT / JiT while retaining their original multi-step sampling procedures.

ELT: Elastic Looped Transformers for Visual Generation

ELT repeatedly applies shared transformer layers within each generation step and distills the full-loop prediction into intermediate exits, allowing a single model to adjust its loop count to the available budget; at matched inference compute, it achieves ImageNet FID 2.0 with 111M parameters, compared with 446M for the corresponding MaskGIT-XL baseline.

RubricRL: Simple Generalizable Rewards for Text-to-Image Generation

RubricRL uses GPT-o4-mini to construct ten inspectable visual criteria per prompt and trains generators with their mean binary satisfaction score, combining this reward with hybrid rollout sampling to improve Phi3's GenEval overall score from 0.7624 to 0.8468 while transferring the same reward interface to diffusion models.

SimFlow: Simplified and End-to-End Training of Latent Normalizing Flows

SimFlow fixes VAE posterior variance to a nonzero constant so that the encoder, decoder, and latent normalizing flow jointly learn generation-friendly representations, achieving 2.15 gFID after 160 epochs on ImageNet 256×256 class-conditional generation and 1.91 with REPA-E.

The Telephone Game: Evaluating Semantic Drift in Unified Models

The paper evaluates unified models through a telephone game that alternates visual understanding and image generation, using embedding-based semantic retention and object-constraint preservation to expose accumulated errors missed by single-pass evaluation; BAGEL achieves the highest full-benchmark MCDavg (0.4414) and MGG (78.2) among seven systems.

Think in Strokes, Not Pixels: Process-Driven Image Generation via Interleaved Reasoning

STIR constructs progressive drawing trajectories with scene graphs and supervises both textual conflicts and image-instruction misalignment, teaching BAGEL-7B to alternate planning, sketching, inspection, and refinement while improving GenEval from 0.79 to 0.83 and WISE from 0.70 to 0.76.

WeEdit: A Dataset, Benchmark and Glyph-Guided Framework for Text-centric Image Editing

WeEdit combines controlled text-editing data, glyph images planned by a vision-language model, and two-stage post-training to raise Qwen-Image-Edit-2509's overall instruction adherence from 3.49 to 7.47 on its bilingual benchmark, although reasoning-based edits remain substantially harder than ordinary text operations.

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.

Curvature-Adaptive Consistency Flow Matching: Autonomous Trajectory Optimization via Reinforcement Learning

CACFM reformulates consistency distillation as a dynamic decision-making process, using a lightweight RL agent to automatically detect and prioritize training on high-curvature bottleneck stages of the PF-ODE trajectory. Combined with flow-adapted DMD and adversarial consistency loss, it achieves state-of-the-art performance with 4-step inference on FLUX and SDXL, outperforming FLUX-schnell by more than 2 FID points.

Early Estimation of Language to Latent Alignment in Diffusion Models

NoisyCLIP fine-tunes the visual encoder of a dual-encoder model on moderately noisy diffusion states to assess candidate-prompt alignment early, reducing the aggregate Best-of-6 denoising budget from 300 to 150 steps while limiting the VQAScore loss against the full-generation CLIP baseline to approximately 2%, as reported in the paper.

EditHF-1M: A Million-Scale Rich Human Preference Feedback for Image Editing

EditHF-1M trains the unified evaluator EditHF using three-dimensional human rankings and scores over 1.01M edited images, then uses its selected preference pairs to optimize Qwen-Image-Edit, raising the editor's human overall score on this dataset from 75.39 to 84.26.

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.

Finite Difference Flow Optimization for RL Post-Training of Text-to-Image Models

FDFO constructs a reward-weighted endpoint difference from two stochastic generation trajectories sharing initial noise and uses it to guide updates along both trajectories, raising the best OneIG-Bench prompt alignment score from Flow-GRPO's 79.56 to 84.25 on Stable Diffusion 3.5 Medium while substantially accelerating progress toward a specified reward level.

GRAN-TED: Generating Robust, Aligned, and Nuanced Text Embedding for Diffusion Models

GRAN-TED uses the text-only TED-6K benchmark to guide encoder selection, then improves Qwen3-VL-8B through visual-semantic fine-tuning and learn-then-freeze layer-wise feature weighting, increasing GenAI-Bench scores over the original Norm-Avg baseline by 1.24 for text-to-image and 2.39 for text-to-video generation.

Personalized Reward Modeling for Text-to-Image Generation

PIGReward explains a user's historical image choices, induces personalized evaluation dimensions, and compares new images dimension by dimension, achieving 85.85% acc w/o tie on PIGBench while also using these preferences for prompt optimization.

PixelU: A U-Shaped Transformer for Efficient End-to-End Pixel Diffusion

PixelU combines shallow-to-deep skip connections for spatial detail with single-stage, constant-channel downsampling for global semantics under clean-image prediction, reaching FID 1.63 at 136.9 GFLOPs on ImageNet 256×256, improving the quality-compute trade-off over JiT-G's FID 1.82 at 383 GFLOPs.

REGLUE Your Latents with Global and Local Semantics for Entangled Diffusion

REGLUE preserves spatial semantics from multilayer visual features with a lightweight convolutional compressor and jointly diffuses them with VAE latents and a global token, reducing FID from 33.0 to 12.9 on ImageNet 256×256 with SiT-B/2 at 400K steps without CFG.

Representation Alignment for Just Image Transformers is not Easier than You Think

PixelREPA addresses the loss of fine-detail diversity caused by REPA in high-resolution pixel-space diffusion by adding partial token masking and a shallow Transformer only to the training-time alignment branch, reducing JiT-B/16 FID from 3.66 to 3.17 on ImageNet 256×256 and reporting over 2× faster convergence measured in training epochs.

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.

Training-Free Refinement of Flow Matching with Divergence-based Sampling

Flow Divergence Sampler (FDS) uses velocity-field divergence to select a less locally ambiguous state from the current state and random neighbors before selected solver steps, requiring no retraining and reducing ImageNet JiT-L/16 + Heun FID from 2.713 to 2.496, versus 2.886 for a wall-clock-matched baseline with more steps.

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.

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.

CGCE: Classifier-Guided Concept Erasure in Generative Models

CGCE places concept erasure before the generator: a learned concept-conditioned classifier applies weighted-gradient refinement only to flagged text embeddings, reducing MMAD attack success on SD-v1.4 from 63.60% to 1.90% without changing generator weights while retaining general generation utility close to the original model.

DreamLite: A Lightweight On-Device Unified Model for Image Generation and Editing

DreamLite combines spatial concatenation and task-progressive training to support generation and editing with one 0.39B denoising backbone, achieving GenEval 0.72 and ImgEdit 4.11 in the full model, while its distilled 4-step version trades some quality for approximately 2.7-second generation and 3.0-second editing on iPhone 17 Pro.

From Sparse to Dense: Multi-View GRPO for Flow Models via Augmented Condition Space

MV-GRPO compares the same generated image group under several semantically adjacent text conditions and reuses its denoising trajectories to update a flow model, raising HPS-v3 from Flow-GRPO's 0.147 to 0.155 on Flux.1-dev with HPS-v3-only training, without improving every text-alignment metric.

Layout-Conditioned Autoregressive Text-to-Image Generation via Structured Masking

SMARLI proposes a structured masking strategy to explicitly control the attention interaction among three types of tokens (text prompts, layout, and image) in an autoregressive Transformer, achieving layout-controlled image generation for AR models without modifying the model architecture. Furthermore, it introduces GRPO post-training to AR layout-to-image generation for the first time, designing a composite layout reward containing VQA, mIoU, and precision, and outperforms most diffusion SOTA models on LayoutSAM-Eval and OverLayBench with only 1.3B parameters.

OpenSubject: Leveraging Video-Derived Identity and Diversity Priors for Subject-driven Image Generation and Manipulation

OpenSubject treats video as the supervision source for subject-driven modelling: cross-frame category consensus plus max-distance DINOv2 pairing mines identity-consistent yet appearance-diverse frame pairs from real videos, and mask-guided outpainting / inpainting synthesizes only the hard inputs while real frames always remain the targets. This yields a 2.5M-sample (4.35M-image) corpus and OSBench, a benchmark covering four generation/manipulation sub-tasks; fine-tuning OmniGen2 and Qwen-Image-Edit-2509 on it raises their average scores by +0.60 and +1.23.

TexTailor: Inference-Time Textual Guidance Tailoring for Multimodal Diffusion Transformers

TexTailor first probes how much each block of an MMDiT model (SD3.5, FLUX, Qwen-Image) depends on the textual condition using three operations — removing, disabling, and enhancing — then fixes a high-signal block range per model and, at inference time, applies a masked amplification of the textual hidden states (default \(\lambda=1.5\)) on only those 4–5 blocks, improving SD3.5's T2I-CompBench Color/Shape by roughly 8%/12% and GenEval overall from 66.42% to 71.63% in a fully training-free manner, with the same block signal reused for precise editing and block-skipping acceleration.

To Erase, or Not to Erase: Robust Training-Free Concept Erasure with Preservation aware Adaptive Ranked Subspace Expansion

PARSE replaces human/LLM-curated static concept banks with a retrieval problem posed to the diffusion model itself — classifier-free-guidance steering fields rank an erase/retain concept bank over the frozen text vocabulary — and then performs training-free editing via a closed-form value-space projection that orthogonalizes the erase basis against the retain basis, plus text-inversion-triggered adaptive subspace expansion; on SD 1.4 it attains the lowest ASR and near-lossless FID/CLIP across NSFW, artistic style, and object erasure.

Vision Bridge Transformer at Scale

ViBT replaces noise-to-target conditional generation with source-to-target Brownian bridge transport, using stabilized velocity matching to adapt large models; its 20B image-editing experiment reports a 2.3-fold inference speedup while the average ImgEdit score decreases from 3.90 to 3.76.

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.

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.

ConceptWeaver: Weaving Disentangled Concepts with Flow

The paper first introduces a differential probing technique showing that concept formation in flow models follows a three-stage "Blueprint–Instantiation–Refinement" schedule, in which content concepts peak in both intensity and spatial localization and become naturally disentangled at mid-stage; guided by this, ConceptWeaver learns low-rank semantic offsets from a single reference image and optimizes them only within each concept's peak stage, then injects them stage-awarely via ConceptWeaver Guidance (CWG), outperforming TokenVerse, MS-Diffusion and DreamBooth on single- and multi-concept image generation (multi-concept S5.5GPT 4.14 vs 3.08/2.98/2.47).

Contrastive Conditional–Unconditional Alignment for Long-tailed Diffusion Model

For class-conditional diffusion models trained on long-tailed data, this paper leaves the architecture untouched and adds two complementary regularizers to the training objective only — an unsupervised contrastive loss with negative samples only that pushes apart the intermediate latents of the unconditional branch, plus an alignment loss that matches the conditional and unconditional noise predictions at large timesteps (the initial stage of denoising) — thereby transferring the data advantage of head classes to tail classes without any inference overhead, lowering FID from 19.9 to 15.1 and tail-class FID from 33.9 to 22.5 on ImageNet-LT 256×256.

DAPS++: Rethinking Diffusion Inverse Problems with Decoupled Posterior Annealing

DAPS++ uses two empirical measurements (the gradient ratio and the gradient inner product) to show that the diffusion prior in inverse-problem solvers actually acts only as a warm initializer, and on that basis fully separates diffusion-based initialization from likelihood-driven Langevin refinement; across super-resolution, inpainting, deblurring and HDR on FFHQ / ImageNet it matches or beats DAPS-1K with roughly 1/10 to 1/20 of the NFEs.

Direct Diffusion Score Preference Optimization via Stepwise Contrastive Policy-Pair Supervision

DDSPO replaces preference targets induced from terminal images with predictions from winning and losing denoising policies, improving SD-1.4 GenEval from D-DPO's 0.4841 to 0.5045 on identical synthetic preference data while supporting policy-pair construction without human preference annotations.

DSH-Bench: A Difficulty- and Scenario-Aware Benchmark with Hierarchical Subject Taxonomy for Subject-Driven Text-to-Image Generation

DSH-Bench samples 459 subjects across 58 fine-grained categories via a three-level hierarchical taxonomy, slices the test cases along two axes (three subject-difficulty tiers x six prompt scenarios) into 5,508 prompts, and scores subject preservation with SICS, an indicator distilled from GPT-4o annotations; it turns subject-driven T2I evaluation from a single aggregate number into a diagnostic report broken down by dimension, difficulty, and scenario, and across 19 models it shows that no method is robust on all categories and that every method degrades on hard subjects.

Glance: Accelerating Diffusion Models with 1 Sample

Glance freezes a pretrained image generator and trains two LoRA adapters on one image-text sample for early semantics and late details, using nonuniform step reduction for the paper's claimed fivefold acceleration, but without preserving text rendering losslessly.

Locality-Aware Continual Unlearning for Diffusion Models

The paper traces the collapse of one-shot concept unlearning methods after only 3-5 sequential deletions to coarse mapping targets, unprotected semantic neighbours, and text-space proxies, and proposes LACU: a single model-aware score-prediction distance drives both a per-prompt nearest safe mapping target and a replay set restricted to the forget concept's nearest neighbours in the model's own representation, which together with teacher-student distillation and ℓ2 parameter regularization holds unlearning accuracy at 0.90 over 10 sequential steps on SD v1.5 while lifting related retention from the 0.34-0.62 baseline range to 0.72.

Momentum Guidance: Plug-and-Play Guidance for Flow Models

The paper proposes Momentum Guidance (MG): it treats an exponential moving average of past model velocities along the sampling trajectory as a "smoother reference branch" and extrapolates the current velocity away from it, giving flow models the same sharpening effect as CFG without any extra network evaluation, auxiliary model, or unconditional branch — and it stacks directly on top of CFG.

Multi-dimensional Preference Alignment by Conditioning Reward Itself

MCDPO conditions a single diffusion model on per-axis preference outcomes so that conflicting image pairs provide directionally explicit supervision and enable inference-time control; on PickV2, the six-metric average win rate for SD1.5 increases from 66.74% with Diffusion-DPO to 81.47%.

OrthoEraser: Coupled-Neuron Orthogonal Projection for Concept Erasure

OrthoEraser identifies sensitive features and the benign features affected by their removal, then subtracts only the sensitive direction orthogonal to a protected subspace, reducing SD1.4's total NudeNet detections on I2P from 646 to 5 while retaining near-original text alignment and output distributions on COCO prompts.

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.

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.

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.

Correlation-Weighted Multi-Reward Optimization for Compositional Generation

To fix a key bottleneck of multi-concept compositional generation — naive aggregation of concept-wise rewards lets easy concepts dominate optimization while genuinely conflicting concepts get diluted — CMO computes a Pearson correlation matrix over concept rewards inside a sample group, estimates each concept's difficulty from its average correlation with the others, and reweights conflicting concepts by \(\text{softmax}(1-\alpha)\); plugged into a MixGRPO + GDPO RL recipe for fine-tuning SD3.5 and FLUX.1-dev, it delivers consistent gains on ConceptMix, GenEval 2 and T2I-CompBench.

Ctrl-Z Sampling: Scaling Diffusion Sampling with Controlled Random Zigzag Explorations

Ctrl-Z Sampling responds to stalled diffusion reward scores by randomly returning to noisier states and progressively deepening its search, matching or exceeding fixed-depth search on several quality measures with fewer denoiser calls, although not every metric improves.

Decoupled Illumination Priors for Spatially Controllable Multi-View Indoor Scene Relighting

Lume-Palette separates four-direction lighting appearances learned from real photographs from spatial lighting distributions supplied by coarse geometry, then coordinates multi-view generation through shared latent states, achieving 24.453 dB PSNR and 0.0090 MV MSE on the evaluated synthetic scenes.

Decoupling Complexity from Scale in Latent Diffusion Model

DCS-LDM uses latent levels to control content detail and a decoder to control output scale, enabling progressive generation through causal hierarchies and asynchronous denoising; its single-level 512px configuration reduces the end-to-end latency in Table 4 from DiT's 6743ms to 1943ms, at the cost of higher decoder memory.

Entropy-Controlled Flow Matching

Entropy-Controlled Flow Matching (ECFM) introduces an entropy rate constraint \(\frac{d}{dt}\mathcal{H}(\mu_t) \geq -\lambda\) to the standard flow matching objective. It theoretically demonstrates that this constraint is equivalent to the Schrödinger Bridge, provides provable mode-coverage certificates, and \(\Gamma\)-converges to classical optimal transport as \(\lambda \to 0\). As a purely theoretical paper in the field of flow matching, it establishes the mathematical foundations of constrained variational principles and entropy-controlled transport.

EruDiff: Refactoring Knowledge in Diffusion Models for Advanced Text-to-Image Synthesis

EruDiff aligns implicit-prompt generation with an explicit-prompt reference distribution and corrects its errors using failed generations, raising FLUX.1-dev's overall WISE WiScore from 0.50 to 0.64 without requiring an additional prompt rewriter at inference time.

FED-Bench: A Cross-Granular Benchmark for Disentangled Evaluation of Facial Expression Editing

FED-Bench builds 747 (source image / editing instruction / ground-truth) triplets with a cascaded pipeline — generate candidates, verify expressions via three-polarity recognition, rank by dual-anchor fidelity preservation, then human re-check — and proposes FED-Score, which disentangles evaluation into three orthogonal dimensions (Fidelity, Alignment, Relative Expression Gain) combined multiplicatively, suppressing both lazy editing and overfit editing across 18 editing models.

Flash-BoN: Instant Drafts for Inference-Time Scaling in Diffusion Models

Flash-BoN spends generation compute on many inexpensive drafts, selects one through multi-stage comparisons, and resumes it with full computation, outperforming full-generation Best-of-N under wall-clock evaluation, with Wan 14B improving from 0.60 to 0.68 normalized AUC on GenAI-Bench.

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.

GIDE: Unlocking Diffusion LLMs for Precise Training-Free Image Editing

GIDE is the first framework to design a noise-inversion mechanism in the discrete token space of diffusion large language models (DLLMs), decomposing image editing into grounding, inversion, and refinement stages to achieve precise localized editing from point, box, or text instructions in a completely training-free manner, accompanied by the 805-case GIDE-Bench.

Histogram-constrained Image Generation

This paper proposes HIG, which leverages Optimal Transport (OT) during the diffusion sampling process to explicitly transform intermediate predictions into a distribution that precisely matches a target histogram. This achieves precise, interpretable, and training-free controllable generation at the histogram granularity, and is extended to high-capacity information embedding and latent space histogram matching.

HorizonRelight: Relighting Long-horizon Videos Consistently via Diffusion Transformers

The paper recasts long-horizon video relighting from "independent short-clip generation per chunk" into "chunk-by-chunk relay of the relit target-domain latent state": training with arbitrary temporal masks teaches the model to continue from partial target-domain context, inference hands the tail latent frames of the previous chunk to the next chunk as anchors, and an externally generated relit starter frame warm-starts the first chunk, which together suppress chunk-boundary seams and cross-chunk appearance drift.

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).

OmniColor: A Unified Framework for Multi-modal Lineart Colorization

OmniColor chooses condition encoders according to whether pixel correspondence is required, preserves lineart boundaries through dense feature supervision, and combines history filtering with semantic gating for sequential colorization, reducing per-sample inference time from 452 seconds for the L+T+H ablation baseline to 152 seconds for the full model.

PhyEditBench: A Real-World Multi-Stage Benchmark for Physics-Aware Image Editing

PhyEditBench is the first benchmark specifically designed to evaluate the physical reasoning capabilities of instruction-guided image editing models, encompassing 4 major categories and 12 subcategories of real-world physical scenarios and anti-physics instances. Concurrently, a training-free method, PhyWorld, is proposed, which leverages video generation models as physical reasoning engines alongside test-time scaling and latent reduction strategies, outperforming most open-source editing models with only 5B parameters.

PIPBench: A Profile-Inclusive Framework for Personalized Image Generation Evaluation

PIPBench incorporates psychological and lifestyle profiles into preference data construction and evaluation, comparing personalized image generation on 1,369 test cases and finding that strong VLM-based textual conditioning fusion outperforms the tested direct-reference approaches while still trading instruction fidelity for preference alignment.

Scaling Multi-Reference Image Generation with Dynamic Reward Optimization

The paper first introduces OmniRef-Bench, covering five reference types and 2-7 reference images per instance, exposing how sharply open-source models degrade as the number of mixed-type references grows; it then proposes DyRef, a two-stage framework that pairs SFT cold-start with two additions on top of Flow-GRPO — DRS, which widens intra-group reward gaps, and DAR, which inversely reweights groups by their mean reward — lifting open-source backbones to a level comparable with the closed-source Nano Banana Pro on complex multi-reference generation.

Self-transcendence: Is External Feature Guidance Indispensable for Accelerating Diffusion Transformer Training?

Self-Transcendence first trains an internal teacher with clean VAE latents, strengthens its semantics using feature-level classifier-free guidance, and uses the frozen teacher to supervise a diffusion model trained from scratch; on ImageNet 256 without sampling CFG, SiT-XL/2 improves from FID 17.63 to 7.51 after 80 student epochs, with teacher warm-up charged separately.

Solving Diffusion Inverse Problems with Restart Posterior Sampling

RePS extends "restart sampling" from unconditional diffusion (piecewise ODE integration plus periodic large noise injections that wipe out accumulated approximation error) to posterior inference: each round runs a few measurement-conditioned ODE steps to push the sample down to the clean regime, solving a small MAP sub-problem in \(x_0\) space at every step (so no gradient ever back-propagates through the score network), then re-noises the sample back to an annealed restart level and starts again — delivering the strongest overall results on 5 linear and 3 non-linear inverse problems over FFHQ/ImageNet, ahead of DAPS, DPS, DDRM, DDNM and DiffPIR.

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%).

Trajectory Forcing: Structure-First Generation with Controllable Semantic Trajectories

Trajectory Forcing (TF) rewrites the opaque denoising trajectory of diffusion/flow models into an explicit semantic coarse-to-fine path (object/background → parts → subparts → detail), where every level is rendered into a viewable image by a shared RAE decoder and can be edited mid-generation before the model continues; each level costs a single network evaluation (L=4 NFE), reaching FID 1.98 / IS 261.6 on class-conditional ImageNet 256×256 after only 80 training epochs.

UniTranslator: A Unified Multi-modal Framework for End-to-end In-Image Machine Translation

UniTranslator proposes a unified multimodal framework to address In-Image Machine Translation (IIMT). It uses an Understanding-Generation Alignment Module (UGAM) to alleviate the supervision conflict between translation prediction and visual rendering at the representation level, and a Spatial Mask Decoder (SMD) to provide pixel-level text region localization, achieving SOTA performance on three benchmarks: Translatotron-V, IIMT30k, and PRIM.

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.

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.

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.

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.

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.

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.

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.

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.

Beyond Fixed Luminance: Towards Panchromatic and Orthochromatic Image Colorization

Addressing the fundamental inability of conventional Lab-based colorization systems to alter brightness—which causes severe degradation on historical red-insensitive orthochromatic photography—this paper proposes a luminance-agnostic full-RGB colorization framework built on a foundation image-editing model (FLUX.2-klein) and trained with a mixed grayscale objective.

Beyond the Black Box: Identifiable Interpretation and Control in Generative Models via Causal Minimality

This paper introduces a theoretical framework for hierarchical selection models grounded in the causal minimality principle, mathematically establishing component-wise identifiability of latent representations in generative models under sparsity constraints and demonstrating grounded hierarchical concept graph extraction, multi-level editing, and state-of-the-art concept unlearning on text-to-image diffusion models.

Beyond Time Shifts: Adapting Omni-LLM as a Reference-Free Evaluator for Generative Audio-Visual Models

To tackle the failure of conventional temporal-shift metrics under authentic generative structural hallucinations, this paper introduces the SynthSync benchmark and adapts an Omni-LLM via continuous latent projection and Real-Valued Group Relative Policy Optimization (R-GRPO) into a high-precision, reference-free synchrony evaluator.

BeyondMasks: Evaluating Causal and Physical Consistency in Video Object Removal

Addressing the limitation of conventional video object removal protocols that evaluate only local inpainting fidelity while ignoring physical side effects like shadows and reflections, BeyondMasks introduces a temporally aligned paired benchmark (180 sequences) and a structured VLM-based evaluation protocol CORE that reveals a systematic gap between pixel-level fidelity and true causal scene restoration.

BiSLW: Bi-Spectral Latent Watermarking for Generative Diffusion Models

Addressing copyright attribution and regeneration-stripping vulnerabilities in generative diffusion models, BiSLW losslessly decomposes latent tensors into complementary low-frequency semantic and high-frequency textural bands via channel-wise 2D-DCT, injecting aligned watermark identities prior to VAE decoding with cross-band consistency constraints to achieve 37.40 dB PSNR and superior regeneration robustness with just 1 ms overhead.

Breaking High Confidence: Practical Face Impersonation under High-Security Thresholds

Addressing critical high-security scenarios such as banking eKYC and airport identity checks (FMR = \(10^{-6}\) or commercial API confidence score 99), this paper systematically analyzes error accumulation in score-based black-box impersonation across metric distortion, subspace projection, and template inversion, proposing a correction matrix, PCA-based orthogonal face set, and SPNet to achieve over 92% attack success rate against AWS Rekognition at a confidence threshold of 99 within 100 non-adaptive queries.

Bridging Online and Offline Handwriting via Differentiable Physical Rendering

This paper introduces a compact six-parameter physical brush model and an analytical differentiable renderer that bridge online stroke kinematics and offline pixel appearance, enabling joint dual-modal handwriting generation and zero-shot robotic calligraphy execution.

Bridging Vision and Language Concepts through Optimal Transport Semantic Flow

Addressing fine-grained localization loss and cross-modal geometric distortion caused by fixed embeddings and global cosine similarity in traditional vision-language CBMs, OTF-CBM adapts cross-modal costs via inverse optimal transport, captures many-to-one matching and background suppression via unbalanced optimal transport, and evaluates instantaneous flow matching velocities at the midpoint for accurate concept activation without numerical ODE integration.

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).

ColorFM: An Optimization-to-Learning Framework for Color Transfer via Flow Matching

ColorFM introduces a two-tier optimization-to-learning framework where instance-level optimization (ColorFM-O) constructs quasi-linear velocity fields via hierarchical color coupling and semantic priors to curate large-scale paired data, and a feed-forward model (ColorFM-L) predicts flow parameters for 16ms bidirectional linearized transport.

Conditional Flow Matching for Visually-Guided Acoustic Highlighting

Reframes visually-guided acoustic highlighting as a generative Conditional Flow Matching (CFM) problem, introducing a trajectory rollout loss to suppress compounding drift and an early cross-modal adapter in the vision encoder for explicit source selection.

Consistent Feature Transport for Image Relighting

Reformulating image relighting as an illumination feature transport problem, this paper proposes Consistent Feature Transport (CFT) under rectified flow, using the parallelogram law and cross-instance illumination trajectory supervision to isolate lighting changes from identity and geometry, establishing new state-of-the-art results on a newly constructed large-scale complex portrait relighting benchmark.

ContextFlow: In-Context Flow Matching for Robot Manipulation

Addressing the quantization errors of action discretization and compounding errors of step-by-step next-token prediction in autoregressive in-context imitation learning, ContextFlow introduces conditional flow matching for continuous action chunk prediction combined with perceiver-style multimodal context compressors, outperforming autoregressive baselines by 35% on unseen LIBERO configurations and matching fine-tuned VLA models without test-time parameter updates.

DeLux: Cross-Modal Local Artifact Restoration in Video Using Neuromorphic Data

DeLux addresses localized pixel saturation and structural information loss in conventional RGB cameras caused by glare, lens flare, and overexposure by introducing a cross-modal local artifact restoration pipeline that leverages neuromorphic event streams (>140 dB dynamic range) to guide targeted inpainting via explicit artifact detection and multimodal fusion, achieving an average MS-SSIM of 0.991 and up to 88% artifact severity reduction in real-world driving footage.

DICT: Data Injection and Contrastive Trajectory Refinement for Conditional Image Generation with Diffusion Models

To eliminate the severe information bottleneck and cumulative sampling errors caused by compressing rich conditioning signals into scalar guidance losses, DICT proposes a unified training-free framework that injects noise-perturbed condition representations during early denoising stages and stabilizes sampling trajectories via contrastive trajectory refinement across adjacent denoising states.

DiffRGD: An Inference-Time Diffusion Guidance Through Riemannian Gradient Descent

Addressing sample quality degradation caused by the destruction of latent Gaussian distributions during inference-time guidance, DiffRGD leverages the polar decomposition of isotropic Gaussians to construct distribution-induced spherical manifolds, achieving high-fidelity conditional generation via Riemannian Gradient Descent.

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.

Discrete Diffusion Bridges for Spatiotemporally Aligned Image Translation and Generation

Discrete Diffusion Bridges (DDB) resolve spatiotemporal misalignment in standard discrete diffusion by introducing a hybrid absorption state as spatial anchors and an information-guided noise schedule to match inference dynamics, achieving superior translation fidelity and robust few-step generation.

DiTailed: Ensuring Visual Object Consistency in Text-Image-to-Image Flow Matching Models

Addressing the frequent degradation of visual object consistency in text-guided image editing, this paper introduces ABO-Edit—a benchmark rendered from 3D assets with degree-level rotation control—and uncovers that rectified flow models inherently encode target predictions in the conditioning latent space at early denoising stages, leading to FlowMirror, a parameter-free multi-scale auxiliary loss that markedly enhances visual fidelity and fine-grained consistency.

DiverseVAR: Balancing Diversity and Quality of Next-Scale Visual Autoregressive Models

DiverseVAR tackles the severe per-prompt diversity collapse in text-conditioned visual autoregressive models (VAR) by introducing condition-annealed noise injection to text embeddings coupled with Scale-Travel—a training-free multi-scale latent rollback and refinement scheme that substantially boosts sample diversity while preserving visual fidelity.

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.

Drift-AR: Single-Step Visual Autoregressive Generation via Anti-Symmetric Drifting

Drift-AR reveals that per-position prediction entropy in continuous-space autoregressive models simultaneously captures speculative draft quality and local generation uncertainty, utilizing causal-normalized entropy alignment and an entropy-parameterized anti-symmetric drifting field to achieve genuine 1-NFE visual decoding with 3.8x to 5.5x end-to-end acceleration.

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.

Dual-Output Multi-Exposure HDR Reconstruction via SDR Fusion and Gain Map Inverse Tone Mapping

DOME-HDR introduces a dual-output multi-exposure HDR reconstruction framework that decomposes the problem into LoRA-adapted latent diffusion SDR fusion and HDR prior-guided gain map estimation (HPGM), concurrently delivering display-ready SDR images and physically accurate HDR radiance within a single pipeline.

DynEval: Holistic Evaluations of T2I Generative Models in the Wild

DynEval introduces a dynamic, human-annotation-free text-to-image evaluation framework that jointly assesses prompt-image alignment and visual quality through curriculum-distilled compact VLMs trained on 250K structured instruction triplets derived from tiered failure mining.

ELDiff: When Evidential Learning Meets Text-to-Image Diffusion

By coupling evidential deep learning (EDL) for uncertainty estimation with Dempster-Shafer theory (DST) for semantic conflict quantification, ELDiff addresses segmentation bias and overlap conflicts in multi-object text-to-image diffusion without any inference-time overhead.

EMAG: Self-Rectifying Diffusion Sampling with Exponential Moving Average Guidance

EMAG is a training-free guidance mechanism for diffusion transformers that temporally smooths self-attention maps via exponential moving average (EMA) alongside statistics-driven adaptive layer selection to construct semantically faithful hard negatives, elevating Human Preference Score (HPS) by +0.54 over CFG on SD3 and composably augmenting orthogonal guidance methods like APG and CADS.

EmbodiedVAE: Disentangled Video VAE for Efficient and Controllable Embodied Manipulation

Addressing the loss of robotic action dynamics caused by generic video VAE compression in embodied manipulation world models, EmbodiedVAE introduces a dual-encoder single-decoder architecture with asymmetric spatio-temporal compression and optimal-transport consistency to disentangle arm motion from background context, achieving a 0.39% compression rate while improving downstream action-conditioned generation by over 2dB PSNR.

EvoTok: A Unified Image Tokenizer via Residual Latent Evolution for Visual Understanding and Generation

EvoTok introduces a residual latent evolution mechanism within a shared latent space via cascaded residual vector quantization, evolving fine-grained pixel details into high-level semantic representations and setting new state-of-the-art benchmarks on ImageNet reconstruction, 9 understanding benchmarks, and complex visual generation benchmarks with only 13M training images.

FDM-MFVT: Few-step Sampling Diffusion Model for Mask-Free Virtual Try-On

Addressing the computational latency and error-prone pre-processing of diffusion-based virtual try-on, FDM-MFVT introduces an Outfit-aware Noise Optimization module (OANO) and an Instruction-driven Try-on module (IDT) that enables high-fidelity mask-free virtual try-on in merely 6 inference steps.

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.

Flow Straight to Reality: Perceptually Consistent Flow Matching for Efficient Image Restoration

PCFlow introduces perceptual consistency into a latent flow matching framework, employing conflict-free gradient projection and an SNR-adaptive schedule to eliminate gradient interference between distortion and perception, achieving high-fidelity restoration using a lightweight convolution-only backbone in just 3 to 5 inference steps.

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.

From Open Loop to Closed Loop: A Test-Time Iterative Optimization Framework for Reference-Consistent Image Generation

This paper reformulates reference-consistent controllable image generation as a closed-loop dynamic tracking control problem, reusing an off-the-shelf feature extractor as both target encoder and feedback sensor while employing a modified residual PID controller at test time to iteratively refine latent control signals without retraining.

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 (Gemini 2.5 Pro) driven three-tier prompting pipeline (diversification + Action Profile disambiguation + In-Context Exemplars anchoring). Combined with a two-stage training strategy (learning features on mixed synthetic and real data, and calibrating on purely real data), it surpasses the strongest long-tailed baselines by 7.0% on K100-LT and 5.1% on UCF-LT, while achieving a 31.9% improvement on RareAct rare actions.

Generalization and Memorization in Rectified Flow

Addressing the underexplored memorization mechanisms in Rectified Flow generative models, this paper proposes the first complexity-calibrated membership inference attack test statistic, theoretically proves that memorization strictly peaks at the ODE trajectory midpoint (\(t=0.5\)) due to state-velocity statistical orthogonality, and introduces symmetric exponential U-shaped timestep sampling to suppress privacy vulnerabilities without degrading generative fidelity.

Generative Action Tell-Tales: Assessing Human Motion in Synthesized Videos

Addressing strong appearance bias and a fundamental lack of temporal kinematic awareness in existing vision encoders and MLLMs, this paper introduces a learned latent manifold of real-world human actions fusing 3D/2D geometry with first-order motion derivatives, proposing reference-free action consistency and temporal coherence metrics that outperform SOTA MLLMs by over 68% in alignment with human judgments on the new TAG-Bench-v0 benchmark.

H-Adapter: Pose-Robust Hairstyle Transfer via Attention-Derived, Source-Aligned Hair Masks

H-Adapter fine-tunes IP-Adapter with a "region-specific loss" such that the adapter injects the reference hairstyle only in the hair region while keeping non-hair regions unchanged. After training, the cross-attention naturally separates hair from non-hair regions, allowing the derivation of a coarse, source-pose-aligned mask to guide a two-stage diffusion inpainting process. This enables accurate hairstyle placement and alignment with the face geometry even when there is a significant pose discrepancy between the source and reference images.

h-Flow: Flexible Flow-based Image Editing via Doob's h-Transform

Addressing the entanglement between source image fidelity and target text alignment in flow-based image editing, this paper bridges deterministic Rectified Flow to Doob's h-Transform via an equivalent SDE, derives closed-form reconstruction guidance and semantic editing signals, and achieves completely decoupled, training-free control via orthogonal velocity projection.

HRDiT: Training-Free High-Resolution Image Generation with Off-the-Shelf Diffusion Transformer Models

Addressing the two fundamental bottlenecks of off-the-shelf Diffusion Transformers (DiTs) when generating high-resolution images in a training-free manner—spatial disorder and quadratic generation latency—HRDiT introduces Spatial Position Alignment (SPA) via token index bundling and sliding to restore positional expressiveness, alongside Head-adaptive Attention Pruning (HAP) via one-pass Taylor expansion and integer linear programming to prune redundant attention computations, achieving superior visual fidelity and substantial inference speedups across 2K, 4K, and 8K resolutions without any model fine-tuning.

Illuminating Unified Multimodal Model for Free-form Interleaved Text-Image Generation

ILLUME-X presents an advanced unified multimodal paradigm supporting free-form, variable-length interleaved text-image generation via decoupled continuous visual encoding, an interleaved attention mask that eliminates cross-step diffusion noise leakage, a decoupled interleaved CFG mechanism, and an automated 100K causal data curation pipeline, matching complex multi-agent systems and substantially outperforming prior unified models on ISG-Bench with a score of 6.26.

Importance-Aware Low-Rank Distillation of Diffusion Transformers

Addressing deployment bottlenecks of large Diffusion Transformers (DiTs), this paper introduces SVDtrunc, a structured parameter compression framework combining block-importance-aware truncated singular value decomposition with modular knowledge distillation, preserving full architectural non-linearities and achieving state-of-the-art generation fidelity at 68% and 57% parameter budgets.

Improving Image-to-Image Translation via a Rectified Flow Reformulation

This paper introduces Image-to-Image Rectified Flow Reformulation (I2I-RFR), a practical plug-in framework that converts standard discriminative I2I regression backbones into continuous-time transport refiners by concatenating noise-corrupted targets, optimizing a time-reweighted pixel loss, and analytically inducing a velocity field, achieving high-fidelity 3-step explicit Euler integration without architectural redesign or distillation.

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.

Inference-Time Scaling of Diffusion Models via Progressive Seed Pruning

Progressive Seed Pruning (PSP) front-loads compute to explore an expanded pool of initial noise seeds, progressively eliminating unpromising candidates via zero-overhead intermediate denoised estimates and black-box rewards to achieve state-of-the-art prompt alignment under a matched compute budget.

InnoText: A Unified Model for Visual Text Generation and Editing

Built on the Flux.1 Fill Diffusion Transformer backbone, InnoText unifies visual text generation and localized editing within a single model via a diptych inpainting paradigm, font size-aware modulation (FSAM), and small-character aware augmentation (SCAA).

InsertAnywhere: Geometrically Grounded and Optics-Aware Video Object Insertion

InsertAnywhere couples 4D scene reconstruction and KNN scene flow tracking with a stop-gradient optics-aware feature alignment strategy, enabling production-grade video object insertion with realistic cast shadows and reflections under dynamic occlusions from a single-frame 3D placement.

InstanceControl: Controllable Complex Image Generation without Instance Labeling

InstanceControl addresses attribute confusion and the heavy burden of manual instance labeling in complex multi-instance controllable image generation by using a VLM to automatically establish text-visual instance correspondences alongside an adaptive mask refinement strategy that dynamically corrects noisy masks during diffusion denoising.

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.

Keep-or-Drop? Adaptive Tokenizer for Compact Video Representation

KATok introduces an adaptive video VAE that jointly learns dynamic token selection via Gumbel-Softmax gating and soft attention masking, coupled with a cascaded mask-prior diffusion strategy to eliminate spatio-temporal misalignment, achieving higher generation quality and several-fold acceleration.

LACON: Training Text-to-Image Model from Uncurated Data

LACON overturns the prevailing filter-first paradigm by repurposing five core quality signals (aesthetic score, watermark probability, clarity, entropy, and luminance) into explicit continuous conditioning vectors via Gaussian-weighted cluster centroids, training on 100% raw uncurated data to delineate quality boundaries and achieve artifact-free, high-fidelity controllable generation.

LensStyle: Learning the Optical Aesthetics for Controllable Stylized Lens Effect Rendering

LensStyle presents the first unified flow matching framework that explicitly learns optical aesthetics, employing a Dual Path Controller to disentangle continuous focal/blur modulation from discrete lens styling (polygonal, donut, cat-eye, and starburst effects) without requiring depth maps at inference time.

LoGAN: Multilingual Font Localization with Generative Agents

LoGAN introduces a VLM-agent-coordinated framework that decomposes multilingual font localization into modular stages—glyph generation, kerning and layout transfer, style fine-tuning, and texture expansion—achieving high-fidelity vector font rendering across 27 languages (including CJK) and outperforming existing text-editing and diffusion baselines.

Logit Refiner: Improving Visual Autoregressive Models via Intra-Scale Dependency Modeling

Visual Autoregressive Models (VAR) suffer from local spatial incoherence due to an implicit mean-field-style parallel decoding rule within scales; Logit Refiner remedies this by sequentially sampling tokens via a lightweight causal transformer over frozen backbone features, boosting generation fidelity with minimal training compute and modest inference overhead.

Look But Don't Touch with Sparse Autoencoders for Unlearning in Diffusion Models

This paper demonstrates that steering latent dimensions in Sparse Autoencoders (SAEs) causes severe out-of-distribution (OOD) activations and visual artifacts in diffusion models, and proposes Patch Embedding Replacement (PER)—a training-free framework that confines SAEs strictly to concept localization and substitutes target patch embeddings with in-distribution activations from non-concept regions.

LooseControlVideo: Directorial Video Control using Spatial Blocking

LooseControlVideo adopts the cinematic concept of spatial blocking by employing coarse, oriented 3D bounding boxes as sparse proxies rendered into depth-modulated normalized object coordinate space (DNOCS), allowing a frozen video DiT backbone to decouple directorial trajectory choreography from fine-grained morphological deformations and achieve high-precision 3D trajectory control and localized video editing.

MapDreamer: Aerial Imagery Conditioned Latent Diffusion For Lane Level Map Generation

MapDreamer formulates lane-level HD map generation from a single aerial image as a latent diffusion process, utilizing a variational autoencoder to jointly compress lane geometry and directed connectivity into a continuous latent space while introducing a lane cardinality module, ghost lane latents, and boundary cross-attention for city-scale vector map synthesis.

MARché: Fast Masked Autoregressive Image Generation with Cache-Aware Attention

To eliminate the severe computational redundancy in masked autoregressive image generation where full bidirectional attention and feed-forward networks are recomputed every step, MARché proposes a training-free framework featuring decoupled cache-aware attention, attention-guided selective KV refresh, and periodic cache calibration, attaining up to a 1.79× speedup with negligible fidelity loss.

MAVIN: Multi-Shot Audio-Visual Generation with Customized Narrative Control

To tackle temporal misalignment, limited controllability over vocal timbre and appearance, and incomplete audio scripting in cinematic filmmaking, MAVIN introduces a multi-agent scripting pipeline, boundary-aware attention, and ID-aware propagation, enabling the first end-to-end multi-shot audio-visual generation framework with precise transition boundaries, dialogue timestamps, and multi-subject identity consistency.

MindFlow: Harmonizing Cognitive Semantics and Acoustic Dynamics for Facial Animation Generation in Dyadic Conversations

Inspired by the neuroscience model of the ventral-dorsal dual-pathway, MindFlow decouples facial animation generation in dyadic conversations into two collaborative pathways: cognitive semantic understanding (Ventral module) and reflexive motion generation (Dorsal module). Using the Chunk-State method, it extracts fine-grained emotional states from audio chunks to guide a flow matching network in generating high-fidelity facial animations.

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.

Multi-Modal Controlled Coherent Motion Generation

Addressing the severe scarcity of aligned multimodal motion data for concurrent walking and talking behaviors, MOCO decouples diffusion denoising across spatial body divisions with per-step iterative refinement and LLM timeline planning, achieving coherent and synchronized full-body avatar motion.

MUSE: Unlocking Timestep as Native Task Steering for One-Step Dense Prediction

MUSE reveals that the native sinusoidal timestep positional embedding in one-step diffusion models can serve as an endogenous, parameter-free task steering switch, achieving manifold decoupling for depth estimation, surface normal estimation, and RGB reconstruction within a single parameter-shared backbone.

NearID: Identity Representation Learning via Near-identity Distractors

Addressing the severe entanglement between object identity and background context in visual foundation models that inflates automated personalization metrics, NearID introduces matched-context near-identity distractors alongside a two-tier hierarchical contrastive loss, tuning only a lightweight MAP head (3.6% parameters) on frozen SigLIP2 to boost sample success rate from 30.74% to 99.17% and substantially enhance human alignment.

NoiseTilt: Noise-Tilted Reverse Kernels for Diffusion Reward Alignment

This paper proposes NTRK (Noise-Tilted Reverse Kernel): during diffusion inference, instead of altering the mean of the reverse kernel, it transforms the reward gradient into a valid Gaussian noise direction via a "whitening operator" and injects it into the noise term. This achieves both "gradient guidance" and "noise compatibility" while sampling only one candidate per step. In aesthetic image generation, with only 25 NFE, it outperforms the strongest baseline at 500 NFE, reducing the computational cost by 20 times.

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.

NumColor: Precise Numeric Color Control in Text-to-Image Generation

Addressing the failure of subword tokenizers that fragment hexadecimal and RGB color codes and break cross-attention binding, NumColor introduces a tokenizer-agnostic Color Token Aggregator (CTA) alongside a 6,707-anchor ColorBook grounded in perceptually uniform CIE Lab space, achieving precise numerical color controllability and zero-shot cross-architecture transfer under frozen diffusion backbones.

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%).

OSOR: One-Step Diffusion Inpainting for Effect-Aware Object Removal

OSOR introduces a one-step diffusion inpainting framework for object removal that combines an occupancy-guided multi-scale discriminator for boundary supervision with a lightweight alpha head for adaptive latent compositing, achieving complete shadow/reflection elimination and mask robustness while outperforming multi-step baselines at 4x to 30x faster inference.

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.

Policy-Based Tuning of Autoregressive Image Models with Instance- and Distribution-Level Rewards

Formulates discrete-token autoregressive image generation as an MDP and applies GRPO policy fine-tuning using a novel EMA-based leave-one-out FID (LOO-FID) distribution-level reward paired with instance-level perceptual rewards and adaptive entropy regularization, resolving diversity collapse while matching or surpassing Classifier-Free Guidance quality to deliver a 2x inference speedup.

Posterior Augmented Flow Matching

Addressing the severe gradient variance and flow collapse caused by one-to-one trajectory supervision in standard flow matching, Posterior Augmented Flow Matching (PAFM) formulates an unbiased expectation over candidate target posteriors via self-normalized importance sampling, cutting gradient variance by 4× and boosting generative fidelity with negligible compute overhead.

PRISM: Latent Composition Consistency for Single-Image Reflection Removal

PRISM observes that mutual coherence between transmission and reflection layers drops ten-fold in pretrained VAE latent space compared to pixel space, reinterpreting reflection removal as a latent linear separation task via single-step flow matching on a FLUX backbone that recovers both layers simultaneously, complemented by cross-sample latent composition consistency and patch-level contrastive separation to achieve SOTA across six benchmarks.

RADIANCE: Relative Adaptive Denoising with IP-Adapter for Novel Concept Enhancement

Addressing attribute omission and semantic drift caused by dominant base-object priors in rare concept generation, RADIANCE introduces a training-free closed-loop feedback framework that monitors latent semantic emergence via CLIP and dynamically applies bidirectional IP-Adapter "restoring forces" to maintain compositional equilibrium.

RankT2I: A Submodular Framework for Discovering Interpretable and Diverse Semantics in Text-to-Image Models

RankT2I is a training-free and model-agnostic semantic discovery framework that leverages multimodal vision-language models to harvest candidate semantics and formulates attribute subset selection as submodular optimization to efficiently uncover diverse, relevant, and reliably editable text-to-image directions.

RawGen: Learning Camera Raw Image Generation

RawGen presents the first diffusion-based generative framework that synthesizes physically consistent, sensor-level linear camera raw images from either text prompts or sRGB images, decoupling unknown and heterogeneous photo-finishing effects through a many-to-one inverse-ISP formulation anchored in canonical CIE XYZ space.

Recolour What Matters: Region-Aware Colour Editing via Token-Level Diffusion

Addressing the limitations where discrete text prompts fail to capture continuous hues and global image conditioning incurs colour drift, ColourCrafter performs token-level fusion of RGB and image features in latent space under a masked Lab-space perceptual loss, enabling fine-grained, structure-preserving local recolouring supported by a newly curated dataset, ColourfulSet.

RefDiT: Local Attribute Guidance in Reference-Based Image Generation

Addressing the failure of prior reference-guided methods to selectively guide generation from local regions in multi-object reference images, RefDiT introduces structured attribute triplet decomposition and a triplet-conditioned LoRA training strategy, enabling high-fidelity local attribute transfer without inference-time architecture overhead.

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.

Rethinking Robust Adversarial Concept Erasure in Diffusion Models

Addressing the fundamental trade-off between robustness and computational efficiency in adversarial concept erasure, S-GRACE harnesses the rich image-text semantic prior of the CLIP text encoder to develop single-sample semantic-guided adversarial optimization and automatic surrogate mapping, robustly erasing sensitive concepts in just 4 minutes while maintaining high image generation utility.

Revisiting Autoregressive Models for Generative Image Classification

This paper revisits the potential of autoregressive (AR) models for generative image classification, identifying that a fixed token order is the key bottleneck limiting AR classifier performance. It proposes using an any-order AR model (RandAR) to average the class-conditional log-likelihoods over multiple random token orders via the Jensen lower bound (order marginalization). This approach comprehensively outperforms diffusion classifiers on ImageNet and various out-of-distribution (OOD) benchmarks while being 25x faster in inference, bringing generative classifiers to a competitive level with DINOv2 for the first time.

RTE-FM-Dehazer: Radiative Transfer Equation Inspired Flow Matching for Real-World Image Dehazing

Addressing the failure of atmospheric scattering models in non-homogeneous real haze and the color drift in generative dehazing, RTE-FM-Dehazer embeds reduced Radiative Transfer Equation (RTE) diffusion-absorption dynamics as a tangent-space regularizer into continuous flow matching, paired with a 50k VLM-synthesized, geometrically aligned dataset (P-HAZE) to achieve artifact-free, high-fidelity real-world dehazing.

ScrollScape: Unlocking 32K Image Generation With Video Diffusion Priors

ScrollScape reformulates the synthesis of ultra-high-resolution images with extreme aspect ratios (e.g., 8:1) as a sequential video panning task, leveraging the temporal consistency of pre-trained video diffusion models as a global structural constraint, combined with Scanning Positional Encoding (ScanPE) and Scrolling Super-Resolution (ScrollSR) to seamlessly scale generation up to 32K resolution while completely eliminating localized repetition and tiling seams.

SegDiff: Segmented Trajectory Diffusion for Consistent and Adaptive Robot Manipulation

SegDiff decomposes demonstrations into keypose-anchored continuous motion segments for long-horizon diffusion trajectory modeling, and introduces a DDIM-inversion-powered Dynamic Temporal Ensembling mechanism that achieves multi-modal consistency and agile real-time adaptation without compounding errors.

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.

Semantically Aligned Gradient-Driven Context-Preserving Image Editing

Addressing the training-time blind spot where instruction-guided generative editors fail to semantically verify their outputs, IABEdit embeds differentiable semantic verification into training by distilling rich descriptors from a frozen VLM via a LoRA-adapted aligner and backpropagating semantic discrepancy gradients into the generative backbone, achieving precise localized editing and background preservation with zero inference-time VLM cost.

Source-Agnostic Image Translation Based on Latent Aware Adaptive Masking

PM-Edit introduces a source-agnostic unsupervised image-to-image translation framework that uses offline prediction discrepancy statistics on the target domain to build a time-dependent dynamic threshold, progressively growing a minimal mask throughout reverse diffusion and guiding inpainting to achieve high-realism and faithful translation without requiring source-domain training.

Spanning Tree Autoregressive Visual Generation

STAR models the image patch grid as a taxicab lattice and employs breadth-first search (BFS) traversal orders of corner-rooted uniform spanning trees as a structured sequence randomization strategy, preserving intrinsic locality and center bias for high sampling fidelity while natively supporting flexible postfix completion for image inpainting via rejection sampling.

Spectral Prior for Reducing Exposure Bias in Diffusion Models

To alleviate exposure bias (distributional shift) in multi-step diffusion sampling that manifests as frequency- and timestep-dependent SNR distortions, this paper proposes Spectral Alignment (SPA)—a method that fits an offline power-law spectral prior and applies lightweight inference-time FFT-based guidance, yielding consistent generation gains across pixel, latent, and flow-matching models with only 3–4% computational overhead.

SR-Edit: Region-Aware Image Editing via Self-Refinement

To tackle unintended non-edit drift and artifact-prone heuristic feature interventions in mask-free image editing, SR-Edit extracts precise edit masks from model predictions via pixel-space self-feedback and enforces non-edit preservation via Doob's h-transform guidance aligned with native sampling dynamics.

Straight-Path Flow Matching for Incomplete Multi-View Clustering

Addressing cross-cluster drift and redundant trajectory overhead in diffusion-based incomplete multi-view clustering caused by isotropic Gaussian noise initialization, this paper proposes a deterministic ODE transport framework based on Straight-Path Flow Matching coupled with cluster-level contrast and entropy-weighted simplex alignment, achieving structure-preserving imputation and robust clustering under single-step integration.

SynVAR: Synergizing Spatial and Semantic Alignment in Visual Autoregressive Model

Addressing the severe cross-scale error propagation and semantic confusion in visual autoregressive (VAR) models during complex compositional generation, SynVAR proposes the first training-free enhancement framework tailored for the VAR paradigm, combining early-stage global spatial guidance and self-attention receptive field constraints with continuous Fourier high-frequency compensation to drastically boost composition accuracy and fine details without retraining.

Test-Time Registers as Global Priors for Tokenized Image Generation

This paper proposes RegToken, a training-free framework that localizes ViT attention sinks using Normalized Feature Norms (NFN), extracts register subspaces via TokenRank, and applies a projection-and-conservation update to repurpose high-norm attention artifacts into global low-frequency priors, significantly improving generation fidelity and alignment for compact 1D token decoders under completely frozen weights.

Text-based Tactile Graphics Generation for the Visually Impaired

This paper introduces the first end-to-end generative graphics pipeline that produces fabrication-ready 2.5D tactile graphics directly from natural language prompts, jointly synthesizing global base geometry, fine-grained tileable tactile textures, and standard-compliant braille for direct SLA 3D printing.

The Illusion of High Utility in Safety Alignment of Text-to-Image Diffusion Models

This paper debunks the illusion that "T2I safety alignment achieves safety without sacrificing capability." It reveals that fine-grained instruction-following capabilities, which are undetected by coarse-grained metrics (FID / CLIPScore), are actually significantly degraded. This degradation is rooted in the "semantic collapse" of the text encoder's embedding space. To address this, the paper proposes SAGE, which recovers structured fidelity with almost no compromise on safety by utilizing two geometric regularization terms (preserving embedding spread + preserving local similarity structures).

Through Van Gogh’s Eyes: Global Style Transfer with Diffusion Model

This paper introduces Global Style Transfer (GST), a many-to-one artistic synthesis paradigm that learns a distribution-level style representation from an artist's full artwork collection via residual h-space modulation under a fixed prompt, complemented by training-free CLIP perceptual guidance for content alignment and style-driven deformation.

Tiled Prompts: Overcoming Prompt Misguidance in Image and Video Super-Resolution

Addressing the prompt misguidance issue in latent tiling super-resolution caused by broadcasting a single coarse global prompt, Tiled Prompts extracts tile-specific localized prompts via vision-language models, theoretically lowering the information-theoretic misguidance bound and substantially boosting perceptual fidelity with minimal compute overhead.

TMI: Text-to-Image Meets Image-to-Image for Complementary Data Synthesis to Boost Long-Tailed Instance Segmentation

TMI introduces a complementary data synthesis framework coupling global text-to-image (T2I) generation for scene diversity with context-aware image-to-image (I2I) editing via VRAIN for verified rare-class augmentation, achieving up to +4.0 overall AP and +9.5 rare-class AP gains on the LVIS benchmark.

TooBad: Backdoor Diffusion Models with Ultra-Low Poison Rate and Imperceptible Trigger

TooBad proposes an end-to-end trigger optimization framework directly embedded in the iterative denoising process of diffusion models under \(L_\infty\) invisibility and sparsity constraints, enabling nearly 100% attack success rates at an ultra-low poison rate of 0.2%~0.5% within only 3~5 fine-tuning epochs while completely evading SOTA backdoor defenses.

Towards In-Context Tone Style Transfer with A Large-Scale Triplet Dataset

Addressing the lack of high-quality paired triplet data and semantic color misalignments in tone style transfer, this paper introduces TST100K—a 100K content-reference-stylized dataset curated via a two-stage tone scorer and aesthetic filtering—and proposes ICTone, a flow-matching DiT framework that treats retouching as in-context visual generation paired with scorer-driven reward feedback learning.

Training-Free Multi-Concept Image Editing

Addressing the linguistic bottleneck where text prompts fail to capture sub-linguistic identity details, this paper introduces Concept Distillation Sampling (CDS), a unified training-free framework that integrates a timestep-ordered, regularised score distillation backbone with dynamic patch-wise LoRA confidence weighting, enabling reference-free and clash-free multi-concept image editing.

TRaM-VSR: Importance-Aware Token Routing and Merging for One-Step Diffusion Video Super-Resolution

Addressing the excessive token computation and temporal flickering of one-step Diffusion Transformer video super-resolution, TRaM-VSR introduces an adaptive token routing and merging framework combining risk-guided offline interval planning, semantic-temporal scoring, two-stream token processing, and route-exit identity restoration, accelerating inference while preserving SOTA perceptual quality and temporal coherence.

UniCSG: Unified High-Fidelity Content-Constrained Style-Driven Generation via Staged Semantic and Frequency Disentanglement

UniCSG introduces a staged semantic and frequency disentanglement framework that builds a robust semantic scaffold via low-pass filtering and conditioning corruption in Stage 1, refines fine-grained textures via latent DoG pyramid supervision in Stage 2, and aligns perceptual quality via pixel-space rewards, substantially suppressing reference-content leakage in both text-guided and reference-guided style transfer.

Unified Backbone Refinement for Diffusion Models via Internal-Latent Analysis

By identifying that abrupt early-stage fluctuations in deep low-noise internal latents directly cause visual artifacts, this paper proposes DUNE, a training-free framework that detects outlier activations via score-consistent EMA tracking and applies backbone-specific suppression across U-Net and DiT models, simultaneously improving visual fidelity and suppressing hallucinations without extra training.

UniGP: Taming Diffusion Transformer for Prior-Preserved Unified Generation and Perception

UniGP builds a lightweight DUGP branch on MMDiT by copying only the image branch and enforcing strict gradient isolation, unifying text-to-image synthesis, any-condition controllable generation, and dense depth/surface-normal estimation within a single model while perfectly preserving generative priors.

Unwarping the Lens: A Physics-Grounded Approach to Video Glasses Removal

To tackle identity drift and temporal flickering in generative eyeglasses removal, this paper transfers large-scale generative priors to a lightweight deterministic network (JFSnet) via rigorous three-stage geometric filtering, physics-based lens refraction/HDR reflection augmentation, and translation equivariance constraints, achieving real-time (27.68 FPS) video restoration with high ocular fidelity.

V-Co: A Closer Look at Visual Representation Alignment via Co-Denoising

V-Co systematically dissects visual co-denoising in a controlled JiT-based pixel diffusion framework and establishes an effective recipe comprising a fully dual-stream backbone, structural semantic-to-pixel CFG masking, a perceptual-drifting hybrid loss, and RMS-based SNR calibration, substantially outperforming prior pixel-space diffusion models on ImageNet-256.

WaterGen: Decoupling Scene and Medium in Underwater Image Generation

WaterGen decouples underwater image generation into degradation-free semantic scene synthesis and physics-grounded conditional decoding, achieving independent and precise control over scene geometry and water medium parameters (attenuation and backscatter) while consistently boosting downstream restoration and segmentation performance as a scalable synthetic data engine.

Wavelet-Guided Semantic Signal Compensation for Inversion-Free Image Editing

To tackle the semantic indistinguishability problem in rectified flow models where early manifold-seeking transport suppresses global attribute modifications, this paper introduces a training-free wavelet-guided semantic compensation framework that extracts co-located low-frequency prompt directions via Haar wavelets and injects them under a quadratic time schedule, markedly boosting global editing capability while preserving background fidelity.

WearWow: Native 2K Multi-Garment Virtual Try-On via Adaptive Token Packing and Preference Alignment

Addressing the dual hurdles of quadratic memory explosion and spectral attenuation of high-frequency fabric textures in ultra-high-resolution multi-garment virtual try-on, WearWow introduces Adaptive 2D Token Packing (ATP) and Multi-dimensional Try-on Reward (MTR) to enable mask-free, photorealistic 2K multi-item synthesis with unprecedented tactile fidelity.

When Cars Have Stereotypes: Auditing Demographic Bias in Objects from Text-to-Image Models

This paper introduces the SODA auditing framework, leveraging controlled prompt generation, two-phase vision-language model attribute discovery, and three standardized metrics (BDS, CDS, VAC) to reveal implicit Middle-aged/White defaults, severe attribute stereotyping (e.g., rose gold laptops for women), and counterproductive homogenization effects under debiasing prompts.

Which Layer Causes Distribution Deviation? Entropy-Guided Adaptive Pruning for Diffusion and Flow Models

To tackle distribution drift and mode collapse during downstream transfer and pruning of Transformer-based diffusion and flow models, this paper proposes Conditional Entropy Deviation (CED) to quantify module-wise damage to the output generative distribution, coupling it with NTK condition numbers and ZiCo gradient proxies into an adaptive progressive pruning framework, EntPruner, which delivers up to 2.22x inference speedup on DiT and SiT with near-lossless generation quality.

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.

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.

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.

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.

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.

Beyond Alignment: A Generative Matching Paradigm via Flow Matching for Zero-Shot Skeleton-Based Action Recognition

ProtoFM reformulates zero-shot skeleton action recognition from conventional heterogeneous skeleton-text alignment into homogeneous skeleton and pseudo-skeleton matching via conditional flow matching (CFM), coupled with a prototype-guided center loss and an online training-free memory bank rectification strategy.

Beyond Linear Shortcuts: Rectifying Diffusion Preference Optimization with Intrinsic Generative Geometry

Addressing the "Linear Shortcut Fallacy" in diffusion Direct Preference Optimization (DPO)—caused by ignoring the intrinsic curvature of the generative trajectory—this paper proposes Trajectory-Consistent Preference Optimization (TCPO), which dynamically anchors on-manifold geometry via Curvature-Adaptive Sampling and fuses semantic preference targets via hyperspherical interpolation, boosting visual quality and achieving a 4.3× training speedup.

Beyond the Boundary: RL-Driven Solution Space Exploration for Blind Face Restoration

LRPO presents the first online reinforcement learning post-training framework for blind face restoration, utilizing diffusion stochasticity for group-based multi-candidate exploration along with a composite reward, GT-guided likelihood regularization, and noise-level advantage assignment to break deterministic trajectory bottlenecks while preserving facial identity.

CHARTSTYLE-100K: A Large-Scale Dataset for Structured Visualization Style Transfer

To overcome data-encoding geometric distortion and reference content leakage in structured visualization stylization, this paper presents ChartForge, a reverse-generation pipeline synthesizing 100K high-fidelity triplets, alongside ChartStyle-Bench and ReChart, a progressively fine-tuned model achieving state-of-the-art element-level style transfer.

CMDer: Controllable Mode Decomposition-Based Single Motion Synthesis with Diffusion

CMDer decomposes a single motion exemplar into multivariate intrinsic frequency modes via Multivariate Variational Mode Decomposition (MVMD) and trains a lightweight diffusion model with a Mode Enhancement Block (MEB) in mode space, enabling frequency-selective control that preserves base rhythmic locomotion while varying fine-grained details.

Co-evolving Representations in Joint Image-Feature Diffusion

Addressing the bottleneck of static, frozen semantic representation spaces in joint image-feature diffusion, this paper introduces Co-evolving Representation Diffusion (CoReDi), which jointly optimizes a lightweight learnable linear projection alongside the diffusion model; equipped with stop-gradient targets, batch normalization, and feature variance regularization to prevent collapse, CoReDi achieves up to a 13x convergence acceleration and superior sample quality in both latent and pixel spaces.

CollectionLoRA: Collecting 50 Effects in 1 LoRA for Deployment

Addressing feature interference, semantic drift, and high routing/storage overhead caused by combining multiple effect and acceleration LoRAs, CollectionLoRA proposes a multi-teacher on-policy distillation framework that seamlessly packs 50 customized visual effects and 8-step fast inference into a single unified LoRA module.

D3F-IR: Dual-Domain Deterministic Flow Matching for Visible-to-Infrared Translation

To resolve the fundamental conflict between macroscopic thermal semantics and microscopic spatial detail reconstruction in visible-to-infrared translation, D3F-IR reformulates the generative process into dual-domain deterministic continuous flows—a latent semantic flow and an image-space pixel flow coordinated via patch-wise semantic modulation—delivering superior structural fidelity and physically plausible thermal radiation synthesis.

Diffusion Image Generation with Explicitly Modeling of Data Manifold Geometry

Addressing the issue that continuous diffusion models diffuse in unbounded Euclidean spaces and ignore data manifold geometry, this paper proposes the Data Manifold-aware Image diffusioN moDel (MIND), which maps discrete tokens onto a compact hyperspherical manifold via a differentiable Soft Top-\(k\) bridge and dynamic multi-stage transition sampling, achieving superior image fidelity with only 130M parameters that outperforms billion-parameter discrete and continuous baselines.

Diffusion to Obfuscation: Time-Adaptive Synthesized Generation Against Gradient Leakage Attacks in Federated Learning

Diff2Obs introduces a client-side diffusion-based gradient obfuscation strategy that generates in-distribution synthesized images with distinct fine-grained details, combining label-agnostic conditioning and cosine time-adaptive annealing to thoroughly thwart both semantic and fine-grained gradient inversion attacks while preserving 98.86% of the model utility.

DisRM: Reward Modeling as Discriminative Prediction

DisRM replaces expensive pairwise human preference annotations and complex multi-metric engineering by training a lightweight discriminator on just 500 unlabeled representative samples to distinguish generator outputs from preferred targets, enabling iterative generator-reward model co-refinement across Best-of-N, SFT, and DPO.

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.

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.

DRPO: Disentangling Demographic Bias from Rewards for Fair Diffusion Alignment

Addressing the issue where preference optimization (PO) amplifies demographic biases encoded in reward models—causing representation collapse in text-to-image diffusion—DRPO extracts demographic directions via linear probing and projects them out via orthogonal complement before reward calculation, achieving high-fidelity generation with restored demographic balance.

DuoFlow: JVP-Free Finite-Difference Mean Flows for One-Step Image Generation

Addressing the computational bottleneck where MeanFlow relies heavily on expensive Jacobian-vector products (JVP), DuoFlow establishes an error-driven JVP-free differential framework that suppresses step-truncation and trajectory-velocity errors, cutting 1-NFE FID by ~28% on ImageNet \(256 \times 256\) from scratch with a \(2.54\times\) end-to-end training speedup.

Dynamic Image Prompt Adapter for Scalable Zero-shot Personalized Text-to-Image Generation

Addressing the core challenges of concept preservation versus prompt following trade-off, loss of fine-grained details, and limited multi-subject scalability in adapter-based zero-shot personalized T2I generation, DynaIP leverages the inherent decoupling learning behavior of MM-DiT dual branches and introduces a Hierarchical Mixture-of-Experts feature fusion module, achieving state-of-the-art single- and multi-subject personalization trained solely on single-subject datasets.

Dynamic World Generation Made Efficient

To overcome the severe multi-hour computational bottleneck and cross-view motion drift caused by sequential autoregressive pipelines, DYNWORLD introduces a Spatial Token Memory that attaches estimated 3D spatial coordinates to intermediate latent tokens in transformer-based video generators, enabling parallel multi-view video generation and decoupled 4D Gaussian reconstruction with an order-of-magnitude speedup (from 3+ hours down to 15–30 minutes).

EmoteGPT: 3D Human Facial Expression from Natural Language Descriptions

EmoteGPT presents the Txt2Emote benchmark featuring explicit feature and implicit situational annotations for 3D facial expressions, and integrates a dedicated <Expr> token within a multimodal large language model to deterministically regress disentangled FLAME expression parameters in real time.

Enlightening Photographic Style Transfer with a Self-Supervised Photographic Embedding

PETAL proposes a dedicated photographic style embedding network trained via self-supervised photographic style augmentation and CIE-Lab histogram guidance, conditioning a lightweight adaptive 5D neural LUT to achieve high-fidelity photographic style transfer without test-time optimization.

EVAR: Edge Visual Autoregressive Models via Principled Pruning

Addressing the severe inference latency on edge devices and the cross-scale error cascading in Visual Autoregressive (VAR) models, EVAR introduces a structured pruning framework combining Pyramid-Aware second-order optimization with Progressive Scale-Aware Distillation (PSAD), pruning VAR-d16 by 40% to 230M parameters while maintaining 3.91 FID and accelerating single-image latency to 277 ms (1.8x speedup) on iOS.

EvDiff: High Quality Video with an Event Camera

EvDiff addresses the computational overhead and paired-data scarcity in event-based video reconstruction by combining Stable Diffusion 3 with a one-step diffusion backbone, a recurrent EvEncoder with Efficient Temporal Fusion, and a 3-stage surrogate training pipeline that distills natural image priors into the event domain, reconstructing high-fidelity, high-dynamic-range chromatic video from monochromatic event streams alone.

Expert Weaving: Marrying Masked AutoRegressive and Diffusion Models for Unified Image Restoration

MoSE identifies complementary task-adaptive priors where Masked Auto-Regressive (MAR) models excel at local structural completion and Diffusion models excel at global texture recovery, introducing a dual-level MoE framework with a DINOv3-guided semantic-aware router and wavelet-based frequency injection that achieves controllable fidelity-perception trade-offs across diverse degradations.

Exposing Implicit Vulnerabilities in Text-to-Image Models via Adversarial Agentic Probing

Addressing the vulnerability of text-to-image models to implicit adversarial prompts that appear benign on the surface yet trigger unsafe images, AdvPIE introduces a black-box, training-free multi-agent red-teaming framework featuring global-relative multimodal feedback and Cumulative Adversarial Decoding, achieving superior attack success rates across open-source and commercial generative models.

FaceArmor: A Universal Facial Image Protection Against Diffusion-Based Manipulations

FaceArmor departs from model-specific white-box defenses by attacking the intrinsic representation feature space of facial images across decoupled structural and semantic tiers, resolving spatial gradient conflicts via surrogate attention heatmaps and PCGrad to achieve universal zero-knowledge protection across unseen diffusion architectures and manipulation paradigms.

Fair and Faithful: A Diffusion-Enhanced Dataset and Hybrid State-Space Mamba for Face Super-Resolution

To tackle the limitations of fixed-path State-Space Models in balancing global facial symmetry with fine-grained local features, as well as the pervasive lack of racial diversity and reproducibility in existing benchmarks, this paper introduces a 70K ControlNet-generated racially diverse dataset (SFHQFSR) and proposes GLASNet—a hybrid state-space network integrating global visual state-space modeling, offset-driven region-adaptive scanning, and channel-spatial-frequency multi-domain refinement that establishes state-of-the-art restoration performance under linear complexity.

FairSteer: Cross-Attention Steering Towards a Fairer Text-Guided Image Generation

EquiSteer (also referred to as FairSteer) is a training-free, per-sample inference-time debiasing framework that employs early-step cross-attention dot-product gating to selectively skip attribute-specific prompts while orthogonalizing and adaptively injecting target attribute steering vectors for neutral prompts, reducing demographic disparity by up to 87% across SD-1.5, SD-2.1, SDXL, and SANA without sacrificing visual quality or alignment.

Fill2SR: Repurposing Inpainting Diffusion Transformers for Real-World Super-Resolution

Fill2SR dispenses with ControlNet-style auxiliary branches and spatial token streams by repurposing the native masked-image slot of an inpainting DiT into a dense low-quality evidence channel under a full-image mask, enabling memory-predictable mixed-resolution training up to native QHD on a single 32GB GPU.

Filterless Snapshot Hyperspectral Imaging: Guided Patch Diffusion Models

This paper proposes a filterless snapshot hyperspectral imaging method based on patch-conditional diffusion models. By using only a single diffractive lens (metalens) and a filterless panchromatic sensor to capture a single grayscale image, it reconstructs an \(H \times W \times 31\) hyperspectral image. The core mechanism is to synchronize the predictions of individual patches during inference via the physical consistency guidance of the optical point spread function (PSF). It outperforms the previous state-of-the-art method by nearly 2.9 dB with 34.6 dB PSNR on ARAD1K, while providing pixel-wise uncertainty estimation highly correlated (\(r = 0.80\)) with reconstruction errors.

FlowInOne: Unifying Multimodal Generation as Image-in, Image-out Flow Matching

By rendering heterogeneous instructions (text descriptions, bounding boxes, visual markers, and force vectors) directly onto an input image canvas, FlowInOne reformulates multimodal generation as a pure image-in, image-out continuous flow matching process within a shared isomorphic latent space.

FlowLess: Controlling Abstract Image Generation

FlowLess introduces a self-supervised framework for controllable abstract image generation from a visual abstraction set (disjoint non-semantic primitives and composition details), leveraging geometric augmentations and per-part modulation to learn abstract degrees of freedom, while using a coverage score to suppress generative clutter and synthesize minimalist compositions.

From smooth to sharp: Frequency-Decoupled Latent Optimization for Realistic Image Generation

DeBaT explicitly decouples low- and high-frequency image bands via discrete wavelet transform and optimizes them using independent encoder-decoder pairs with specialized objectives, resolving the low-frequency bias of standard latent tokenizers and substantially improving fine details and sample realism in latent diffusion generation.

G3AFT: Glance Guided Gradient Aligned Fine-Tuning for Visual Autoregressive Models

Addressing the non-differentiability of discrete token sampling and the prohibitive memory footprint in visual autoregressive modeling, G3AFT establishes the first end-to-end differentiable fine-tuning framework using periodic gradient recording and causal subspace projection to achieve robust reward alignment.

GaINeR: Geometry-Aware Implicit Neural Representation for Image Editing

GaINeR incorporates learnable Gaussian distributions into the Implicit Neural Representation (INR) framework, aggregating neighborhood features via radius-limited Gaussian-weighted KNN to condition an MLP decoder, delivering breakthrough 2D image reconstruction fidelity alongside robust, artifact-free physical deformation and single-image 3D lifting.

Gender Bias in Vision-Language In-Context Learning

This paper systematically investigates how multimodal in-context demonstrations exert a "directional force" that amplifies gender bias in large vision-language models primarily via an asymmetric cross-gender degradation mechanism, and proposes an offline synthetic demonstration image replacement strategy using diffusion models that mitigates bias without degrading caption quality.

Generative Manifold Distillation: Aligning Restoration Trajectories with the Natural Image Prior

Generative Manifold Distillation (GMD) operates under a strictly unsupervised setting with only unannotated low-quality target observations, performing offline orthogonal projection onto the natural image manifold via a frozen 12B foundation model and distilling it into a lightweight restorer with quality-gated filtering and source trajectory anchoring, delivering state-of-the-art out-of-distribution performance with zero added inference latency and zero architectural changes.

Generative Refinement Network for Visual Synthesis

GRN introduces theoretically near-lossless Hierarchical Binary Quantization (HBQ) to resolve the discrete tokenization bottleneck and couples it with a global refinement mechanism inspired by human painting alongside entropy-guided adaptive sampling, eliminating autoregressive error accumulation while achieving complexity-aware adaptive-step visual synthesis.

Geometry-Preserving Image Generation for 6D Object Pose Estimation

To tackle the acute bottleneck of costly real annotations and the synthetic-to-real domain gap in 6D object pose estimation, this paper introduces GenerationPose—a Diffusion Transformer (DiT) framework integrating ControlNetGeom and mask-aware self-attention blocking to convert synthetic 3D CAD/mesh/point-cloud renders into photorealistic pseudo-real images that strictly preserve 6D object pose labels.

GLARE: Towards Generalizable Detection of Latent Diffusion Images with Global-Local Reconstruction Error

By contrasting full-image reconstruction against isolated patch-wise reconstruction within a shared pretrained autoencoder and calibrating with shallow edge density, GLARE provides a completely training-free detector that exploits the elevated context dependence of latent diffusion models with state-of-the-art cross-generator generalization.

GMODiff: One-Step Gain Map Refinement with Diffusion Priors for Efficient HDR Reconstruction

Addressing the limited dynamic range compression, multi-step sampling overhead, and generative hallucinations when applying latent diffusion models to multi-exposure HDR imaging, GMODiff reformulates HDR reconstruction as a degradation-aware gain map refinement task, leveraging regression priors to steer single-step diffusion denoising and decoding to achieve state-of-the-art perceptual quality with a 100× inference speedup.

Hi-DiT: Hybrid Latent-Pixel Diffusion Transformer for Image Generation

Hi-DiT unifies latent-space and pixel-space diffusion inside a single parameter-shared Diffusion Transformer, leveraging temporal denoising heterogeneity to activate a time-gated pixel pathway and hierarchical sub-pixel predictor at low-noise timesteps for state-of-the-art visual fidelity.

Hierarchical Style Aggregation for Versatile Chinese Handwriting Generation

HiSAC explicitly decomposes Chinese characters into radical, component, and stroke structural hierarchies, pairing them with a multi-band frequency-aware style encoder and cross-scale aggregation to unify both online trajectory generation and offline diffusion-based handwriting synthesis.

InstaEdit: Instant Image Editing via Optimized Noise Prediction

Addressing the excessive latency caused by multi-step iterative denoising in instruction-guided image editing, InstaEdit predicts an optimized initial noise state conditioned on both the reference image and editing instructions, enabling models to start from intermediate latent states along the generative trajectory and achieving ~4x end-to-end acceleration with only 4 sampling steps while matching or exceeding 30-step generation quality.

InstantRetouch: Personalized Image Retouching without Test-time Fine-tuning

RefRetouch (titled InstantRetouch) achieves generic personalized image retouching without test-time fine-tuning by coupling a LoRA-adapted Siamese SigLIPv2 encoder with a color-space conditional MLP decoder to disentangle content from retouching style, and introduces retrieval-augmented retouching (RAR) to dynamically aggregate relevant reference styles based on photometric content similarity while generalizing out of the box to photorealistic style transfer.

JSON: Jigsaw Self-play Optimization for Normalizing Flows

Addressing the lack of predefined diffusion dynamics in Normalizing Flows (NFs) and the training collapse induced by standard self-play objectives, this paper proposes JSON—a framework integrating an intrinsic jigsaw puzzle reassembly evaluator with bounded self-play loss to achieve state-of-the-art generation performance without external vision priors.

Jumping the Landing Phase: Noise Variance Matching Enables Accurate Few-Step Inversion

Addressing the systematic underestimation of noise magnitude in DDIM inversion caused by input temporal mismatch, this paper introduces Noise Variance Matching (NVM), which aligns predicted noise variance with a forward reference to jump the costly landing phase and achieve accurate few-step inversion with minimal UNet calls.

LayerVerse: Finding the Sweet Spot for KV-Injection in Training-Free Image Editing

LayerVerse addresses the fundamental trade-off between background preservation and target editability in training-free MM-DiT editing by decoupling KV-injection into masked background protection and sparse global structural anchors, globally optimizing layer allocation via Mixed-Integer Linear Programming (MILP) over pairwise interaction dynamics.

Learning to Corrupt for Better Restoration

To eliminate heuristic corruption strategies that diverge from pretrained diffusion trajectories in one-step image restoration, IAC-IR predicts sample-specific optimal timesteps and noise via intrinsic Gaussian statistical supervision and denoiser consistency, seamlessly integrating them into score distillation to achieve superior one-step perceptual reconstruction.

Learning to Stylize by Learning to Destylize: A Scalable Paradigm for Supervised Style Transfer

This paper introduces a reverse supervised learning paradigm—learning to stylize by learning to destylize—and develops DeStylePipe with a Chain-of-Thought filter to construct DeStyle-350K, a large-scale dataset of 350K pixel-aligned triplets with authentic artistic targets, alongside the balanced evaluation benchmark BCS-Bench.

LINA: Learning INterventions Adaptively for Physical Alignment and Counterfactual Generation in Diffusion Models

Addressing the physical violations and counterfactual generation failures caused by symmetric denoising in diffusion models, LINA models generation via Causal Scene Graphs, predicts prompt-adaptive token/latent interventions via a lightweight module, and reallocates early denoising computation, achieving SOTA physical alignment without model retraining or online MLLM inference.

LISA: Locality-Informed Speculative Decoding for Accelerating Autoregressive Image Generation

Addressing the collapse of acceptance rates in speculative decoding caused by flat, non-discriminative visual token distributions, LISA exploits the local geometry of visual codebooks via locality-informed distillation and geometry-aware soft verification, achieving up to 4.12× lossless acceleration.

LUA: Latent Upscaling Adapter for Diffusion-Based Image Synthesis

Addressing the heavy computational cost of high-resolution diffusion denoising and the artifacts of post-hoc super-resolution, LUA inserts a lightweight feed-forward latent upscaling adapter before the frozen VAE decoder, combining a shared Swin backbone with a three-stage latent-pixel curriculum to achieve native-level 2K/4K synthesis via a single decode with multi-fold acceleration.

MANGO: Unleashing Image Generation Capability of Unified Multimodal Models

To tackle cross-functional interference and residual defects in unified multimodal models during complex compositional generation, MANGO presents Functionality-Oriented Transformers (FoT) to raise the performance floor by disentangling visual understanding and generation branches, alongside Paint-CoT and a multi-task joint training scheme to elevate the ceiling through planning, acting, reflection, and inpainting-based correction.

MeanTalker: Efficient and Expressive Speech-Driven 3D Facial Animation via Geometric-Aware Mean Flow

Tackling the severe latency of diffusion sampling and the over-smoothing of deterministic regression, MeanTalker introduces Mean Flow into speech-driven 3D facial animation via a Dual-Time Denoising Transformer (DTDT) and Geometric-Aware Trajectory Learning (GATL), achieving single-step inference (NFE=1, RTF≈0.004) with superior lip-sync precision and expressive facial dynamics over diffusion SOTA.

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.

MetaPoint: Unlocking Precise Spatial Control in Visual Generation

MetaPoint introduces an architecture-free, single-token spatial interface that directly reuses the native continuous 2D positional encoding of unified multimodal models to represent floating-point coordinates, enabling pixel-accurate multi-object generation and agent-driven interactive editing.

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.

Motion Style Slider: Endpoint-Supervised Continuous Style Control for Human Motion Diffusion

To address the lack of fine-grained continuous intensity control in human motion diffusion, this paper proposes Motion Style Slider, an endpoint-supervised framework that models style direction in a frozen latent embedding space and controls generation via a scalar slider \(\alpha\), leveraging diffusion denoising alongside linearity and monotonicity latent regularizers to achieve smooth, monotonic, and extrapolatable style scaling without requiring intermediate ground-truth motions.

MPO: Single-Stream Policy Optimization for Efficient Text-to-Image Alignment

MPO proposes a group-free, single-stream online reinforcement learning framework (one prompt, one trajectory, one policy update) that eliminates group synchronization latency and avoids intra-group advantage collapse via SDE exploration, Girsanov-drift-adaptive Bayesian value tracking, and global advantage normalization, achieving a 26x wall-clock speedup and a 5x sample-efficiency gain on FLUX.1 Dev.

Occlusion-Resilient Category-Agnostic Pose Estimation with Conditional Flow Matching

FlowCape reformulates Category-Agnostic Pose Estimation (CAPE) as a structured pose transport process driven by conditional flow matching, integrating heatmap-guided initialization to anchor candidate modes and a topology-constrained Riemannian Pose Head with Graph Flow Encoders to achieve state-of-the-art pose recovery under severe occlusion.

OmniPoser: Flexible Human Motion Recovery in the Wild with Masked Flow Matching

OmniPoser unifies heterogeneous and missing sparse sensor signals into a canonical body representation using a cross-modal masking mechanism and a Masked Conditioning Encoder (MCE), achieving full-body SMPL motion recovery via decoupled dual-stream Conditional Flow Matching in a single Euler step at >700 FPS.

On the Diffusibility of High-Dimensional Latents

Addressing the optimization failure of standard velocity prediction in high-dimensional representation autoencoders caused by effective dimensionality collapse after reconstruction finetuning, this paper analyzes how orthogonal noise components disrupt diffusion learning and demonstrates that clean data (\(x_0\)) prediction successfully isolates the low-dimensional signal manifold, enabling high-fidelity visual generation without latent compression.

On-Policy Diffusion Reinforcement Learning Meets Off-Policy Quality Anchoring

DiffusionCompass breaks the self-limiting myopia of on-policy diffusion reinforcement learning by employing off-policy quality anchors as a stabilizing global compass—combining distribution-aware filtering, perceptual anchored rewards, and joint reward renormalization to simultaneously achieve superior reward optimization, enhanced visual diversity, and accelerated convergence across image and video generation tasks.

Online Versatile Incremental Learning: Towards Class and Domain-Agnostic Adaptation at Any Time

To address the realistic scenario of concurrent, continuous, and boundary-free shifts in both classes and domains, the paper formalizes Online Versatile Incremental Learning (Online VIL) and introduces TopFlow, which couples Domain-agnostic Flow Matching (DFM) with Global Topology Preservation (GTP) to enable state-of-the-art anytime and final adaptation without replay memory.

ParaFlow: Parallel Sampling for Flow Matching Models

ParaFlow reformulates the autoregressive ODE sampling of Flow Matching models into a system of Triangular Nonlinear Equations (TNEs) and introduces a lightweight velocity extrapolation scheme via Newton's first-order divided difference, achieving up to a 4.3x wall-clock speedup on a single GPU with fewer total function evaluations (NFEs) than sequential Euler integration.

Parsimonious Flow Matching for Efficient Image Generation

Parsimonious Flow Matching (PFM) replaces the standard isotropic Gaussian latent with a mixture of Gaussians (MoG) geometrically aligned with the multi-modal, low-dimensional data distribution, paired with class-conditional optimal transport and covariance eigenvalue regularization, cutting training iterations by 25%–50% while markedly enhancing low-NFE synthesis quality.

PhysEdit: Physically Consistent Image Editing via Causal Enforcement

Addressing the issue where conventional single-step image editing neglects physical laws—causing floating objects and missing shadows or reflections—PhysEdit reformulates editing into an autoregressive causal evolutionary process, combining single-step teacher consistency regularization with fine-grained physics-aware reinforcement learning to achieve physically consistent edits.

Physics Meets Perception: A Reinforcement Learning Framework for Unpaired Real-World Image Dehazing

Addressing GAN training instability and prohibitive diffusion sampling costs in unpaired real-world dehazing, Dehaze-RL reformulates the task as a one-shot Contextual Bandit with a Physics Hybrid Action Space and optimizes it via GRPO under multi-granularity perceptual and physical reversibility rewards.

POET: Preference Optimization for Enhanced Text-to-Image Generation

POET introduces an automated, SFT-free prompt rewriting framework trained via iterative Direct Preference Optimization with multimodal LLM judge feedback, substantially bridging the distributional gap between underspecified user inputs and frozen text-to-image backbones.

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.

Push–Pull Attentional Anchoring for Diffusion Concept Erasure

Push–Pull Attentional Anchoring (PPAA) regulates weight modification in diffusion cross-attention key-value projections via bounded relative cosine-similarity margins, achieving Pareto-optimal concept removal without corrupting unobserved same-domain peer entities.

RCEdit-500K: Reference Completion for Image-Conditioned Image Editing

Reformulates image-conditioned image editing (ICIE) data curation as a reference-completion problem by synthesizing type-specific reference images for existing high-quality text-guided editing triplets, building RCEdit-500K—the first unified open dataset spanning 477K quadruplets across six edit types.

Rdm: Re-conceptualizing Distribution Matching as a Reward for Diffusion Distillation

This paper re-conceptualizes Distribution Matching Distillation (DMD) as an intrinsic policy reward (\(R_{\mathrm{dm}}\)), and proposes a unified Group Relative Policy Optimization framework (GNDMR) featuring group normalization and adaptive magnitude scaling that eliminates high-noise score variance and seamlessly integrates auxiliary aesthetic rewards for few-step diffusion models.

ReAL: Reference-to-Image (R2I) Aware Latent Diffusion for Image Super-Resolution

ReAL departs from the conventional dependence on text-to-image (T2I) priors and classifier-free guidance (CFG) in super-resolution, establishing a pure reference-to-image (R2I) latent diffusion framework with one-time reference key-value caching and direct self-attention injection, drastically cutting latency and memory while eliminating semantic drift.

Region-Aware Test-Time Scaling for Compositional Image Generation

Addressing the "scaling plateau" of global test-time exploration on complex multi-object and relational compositions, this paper introduces Region-Aware Scaling (RAS) to treat regional decomposition as a dedicated scaling axis, injecting localized guidance only during early denoising to match a 32-sample noise-scaling baseline with just 2 samples.

Rethinking Cross-Spectral Image Generation via Shared-Specific Representation

Addressing semantic misalignment and distorted thermal properties in visible-to-infrared cross-spectral translation, this paper proposes SHASP, which reframes cross-modal synthesis as a latent decoupling and reconstruction process of shared structural and modality-specific spectral features, delivering state-of-the-art structural consistency and thermal fidelity.

Roam2Room: A Unified Floorplan-to-Furnished Framework for Controllable Indoor Scene Generation

Roam2Room (HomeWorld) addresses data scarcity and incoherence in whole-home 3D synthesis by establishing a four-stage pipeline—coupling K-D tree-based floorplan LLMs, dual-view 2D-to-3D roaming in an unfurnished shell, VLM closed-loop reflection, and surface manipulable object placement—to synthesize simulation-ready environments.

RobustRDP: Advancing Reaction Diagram Parsing via Synthetic-to-Real Data Scaling and Robustness-Oriented Training

RobustRDP addresses two core bottlenecks in chemical reaction diagram parsing—expert annotation scarcity and autoregressive error accumulation—by introducing a "sample-render-arrange" synthetic-to-real data scaling pipeline alongside a three-stage progressive training strategy that integrates region guidance, prefix perturbation, and Direct Preference Optimization (DPO), significantly surpassing previous SOTA baselines.

SCALE: Semantic-Calibrated Guidance Enhancement for Prompt-Faithful Diffusion

Addressing the quality–alignment tension caused by coupled linear extrapolation in Classifier-Free Guidance (CFG), SCALE introduces a training-free, orthogonal-invariant enhancement mechanism that selectively scales semantic progress along the prompt axis while strictly preserving orthogonal corrective components, significantly improving prompt adherence with negligible inference latency.

Setting the Stage: Text-Driven Scene-Consistent Image Generation

Addressing the core challenge of visual storytelling where a persistent stage must be established under dynamic camera shifts before placing actors, this paper proposes a paired data construction pipeline leveraging entity removal and image-to-video (I2V) generation alongside a correspondence-guided attention regularization loss, enabling scene-consistent image synthesis under significant viewpoint and composition alterations.

SGP2: Coarse-to-Fine Controllable Multimodal Remote Sensing Image Generation

To tackle semantic conflicts and physical inconsistencies in remote sensing image synthesis, SGP2 establishes a coarse-to-fine generation framework that disentangles global vs. detailed text concepts along Grassmann manifold geodesic trajectories and dynamically routes multi-level visual foundation features constrained by frequency-domain amplitude and phase spectral losses.

SIMON: SImultaneous Multi-Object Navigation

SIMON tackles the critical challenges of attention dispersion, ghosting artifacts, and multiplicative inference latency in multi-object image editing through a training-free framework featuring content-aware attention, multi-object energy guidance, and latent initialization, achieving superior visual fidelity and constant inference efficiency in a single reverse diffusion trajectory.

SpecV: Specification Verification for Robust Unified Multimodal Evaluation

Addressing the severe vulnerability of unified multimodal models (UMMs) to evaluation drift under holistic VLM-as-a-judge scoring, SpecV reformulates open-ended multimodal assessment into atomic binary specification checking via the Specification Verification Protocol (SVP) backed by the multi-model Specification Ensemble and Refinement (SER) pipeline on the 1,200-instance SpecV-Bench, substantially enhancing cross-judge agreement and evaluation reproducibility.

Swap the Right Identity: Spatio-Temporal Preference Optimization for Identity Swapping

To break the trade-off deadlock among identity consistency, background fidelity, and pose stability in character swapping, this paper proposes SIPA, a spatio-temporal identity preference alignment framework that introduces Region-Weighted Preference (RWP) to focus gradients on critical foreground regions and Time-Gated Identity Preference (TIP) to inject ArcFace perceptual penalties exclusively during low-noise steps.

Temporally Stable Generative Illumination with a One-Step Diffusion Model

Addressing slow multi-step sampling and temporal flicker in neural global illumination, this paper proposes a single-step latent diffusion framework conditioning on intrinsic signals and sparse lighting hints via dual adapters, combined with a plug-and-play Temporal VAE decoder to synthesize geometrically aligned, flicker-free indirect illumination at 0.29s per frame.

The Map Is Not the Territory: Embedding-Coverage Blacklists for Safe Diffusion Steering

Addressing the long-standing bottleneck of hand-curated blacklists in trajectory-steered safe text-to-image generation, this work uncovers that blacklist efficacy is governed by embedding-space geometric coverage rather than keyword precision or list size, introducing MAPS to automatically generate and dynamically route high-coverage blacklists via reinforcement learning and farthest-point selection, reducing attack success rates by 25%–52% relatively without model modification.

The Path to Reconciling Quality and Safety Alignment in Text-to-Image Generation

Addressing the debilitating trade-off in text-to-image safety alignment where safety constraints degrade generation aesthetics and instruction fidelity, this paper proposes a unified framework comprising the dual-annotated dataset LibraAlign-100K, the Synergistic Preference Optimization algorithm (T2I-SPO), and the balanced Unified Alignment Score (UAScore), effectively eliminating diverse NSFW concepts while fully preserving generative quality.

TiltDiff: Tilted Weight-Space Diffusion for Neural Network Generation

TiltDiff introduces a performance-tilted latent diffusion framework for neural network weight generation that aligns network parameters into a compact Transformer latent space and biases the diffusion denoising objective with validation accuracy, synthesizing high-performing, robust, diverse, and functionally interpretable model weights.

TOPA: Mitigating Concept Dominance in Diffusion Personalization via Target-Oriented Perturbation Augmentation

Addressing the concept-dominant failure in few-shot diffusion personalization where the customized concept suppresses prompt-specified context, TOPA presents an offline, reference-image-only data augmentation framework that enriches backgrounds via subject-isolation compositing and optimizes additive spatial perturbations to rebalance token-region cross-attention, restoring contextual controllability without altering downstream personalization pipelines.

UF0-6D: Unified Flow-based Zero-Shot 6D Object Pose Estimation without Refinement

UF0-6D reformulates zero-shot 6D object pose estimation as conditional Riemannian flow matching on the \(SE(3)\) manifold, learning an instance-conditioned posterior via geodesic-consistent bridge velocities and symmetry-aware optimal transport to replace coarse-to-refine cascades and iterative render-and-compare loops with a single ODE integration pass.

UltraGen: Efficient Ultra-High-Resolution Image Generation with Hierarchical Local Attention

UltraImageGen addresses the quadratic attention explosion and scarce high-resolution training data in text-to-image diffusion transformers by pairing window-first local window attention with position-scaled low-resolution global guidance and lightweight LoRA joint denoising, delivering state-of-the-art 4K/8K synthesis with over 10x acceleration without requiring any native high-resolution training data.

UNet-Twice: A Simple Structured Reference-based Inpainting Framework

UNet-Twice proposes a simple yet effective framework that reuses the exact same diffusion UNet in two passes per forward step alongside complementary masking and coarse geometric alignment, achieving state-of-the-art reference-based inpainting without any extra branches or cross-attention bridges.

Unifying CNNs and ViTs for Learning-Efficient and Scalable Variational AutoEncoder

TransVAE unifies a shallow CNN front-end for local feature extraction with a deep Transformer backbone for global context modeling, leveraging a pure RoPE strategy, multi-stage overlapping convolutions, and a convolutional feed-forward network to achieve 4× faster convergence, robust zero-shot resolution extrapolation, and predictable scaling from 44M to 2.3B parameters.

UniScale: Arbitrary-Scale Anomaly Generation

Addressing the severe information loss of fine-grained defect details in diffusion latent space caused by compounded spatial downsampling, UniScale proposes an Error-Suppressed Multi-Scale Training (EMT) strategy alongside Generation-then-Fusion Denoising (GFD), enabling high-fidelity industrial anomaly synthesis across arbitrary scales while preserving critical location-aware semantics.

V-HOLD: Stabilizing Flow Trajectories to Rethink the Edit–Preservation Trade-off

Addressing the edit–preservation trade-off in generative flow-based editing, this paper identifies integration trajectory instability as a primary driver of semantic degradation, and proposes V-HOLD, a training-free framework that stabilizes editing dynamics via anchor-step candidate selection and cross-step directional holding, improving semantic preservation while reducing model evaluation costs.

Walk through Paintings : Ego-centric World models from Internet Priors

EgoWM presents an architecture-agnostic conditioning framework that converts off-the-shelf video diffusion models into action-conditioned egocentric world models by injecting compressed, temporally aligned action trajectories directly into the universal denoising timestep pathway, scaling from 3-DoF navigation to 25-DoF humanoid manipulation with high fidelity, extreme OOD generalization into paintings, and introduced the Structural Consistency Score (SCS).

WALL-EVE: World Alignment with Rule Learning in Visual Environments

Addressing the compute-heavy training and physical rule hallucinations of generative video world models, WALL-EVE introduces a training-free neurosymbolic world alignment paradigm that learns compact, executable Python rules from prediction-vs-reality conflicts to calibrate an LLM world model, achieving superior embodied planning under Model-Predictive Control with only 17% of baseline token cost on ALFWorld and Minecraft.

When Distillation Breaks Motion Control: Restoring Generative Trajectories for Fast Video Generators

MotionEcho retains a distilled student's few-step generation backbone and calls a fine-step teacher only when motion errors warrant clean-latent endpoint correction, achieving motion fidelity of 0.931 in 9 seconds with 8 steps on TurboBench, versus 0.876 in 114 seconds for MotionClone on full VideoCrafter2.

When Higher Order Hurts: Pre-Asymptotic Order Collapse in Generative ODE Sampling — A Theory of Discretization–Learning Interaction

Applying backward error analysis reveals that multiplicative interaction terms \(h^k \nabla^k \varepsilon\) between discretization and structured learning error cause higher-order solvers to underperform first-order Euler under low-NFE budgets, yielding a predictive crossover step size \(h^*\) and an adaptive order-selection scheme operating at zero extra NFE.

Zero-Shot Image Personalization from Personas

ZIPP uses natural-language user personas as interpretable preference conditions: an LLM roleplays the persona to rewrite image prompts, while a vision-aligned social-graph encoder mines the personas. It improves personalized generation without per-user fine-tuning, but "zero-shot" refers to the generation interface once a persona is given — it does not mean persona construction is entirely free of user data.

μFlow: Leveraging Average Images for Improving Generalisation of Deepfake Faces Detectors

To tackle the severe performance collapse of deepfake detectors on unseen generators, μFlow introduces a one-class detection framework trained solely on real face images by mapping single-image features into a highly discriminative latent distribution estimated from averaged real images via a normalizing flow, achieving SOTA generalization across 19 unseen GAN and diffusion model architectures.