多模态生成与编辑

Modern video generators routinely fail at physical dynamics: objects float, trajectories violate gravity, contacts vanish. Standard denoising and flow-matching objectives fit visual data distributions but do not explicitly penalize such physical violations. Existing remedies can improve physical consistency, but typically add substantial inference or training cost. Candidate-selection methods generate and score multiple videos, while gradient-based world-model guidance repeatedly decodes and re-encodes intermediate estimates. Generator-internal refinement adds perturbation and re-denoising loops, whereas post-training requires curated data and additional optimization. We propose Off-Manifold Refinement (OMR), an inference-time method that instead injects world-model feedback directly into a single sampling trajectory. During scheduled middle ODE steps, we augment the generator velocity with the gradient of an adapter-space V-JEPA 2.1 surprise energy. This external correction can move the latent away from the uncorrected sampling trajectory and toward regions ranked as more physically plausible by the frozen predictor, after which the generator continues rendering from the corrected state. A small trained latent-to-embedding adapter keeps the gradient tractable at inference, and both the video generator and the world model remain frozen. On our fixed 400-prompt VideoPhy-2 detailed subset, OMR lifts the joint Semantic-Adherence-and-Physical-Commonsense metric from 47.0% to 52.0% (+5.0pp absolute, +10.6% relative) over the base Wan2.2-T2V-A14B sampler. On a separate fixed 50-prompt efficiency subset, it requires $1.71 \times$ the base runtime rather than the multiplicative cost of reward/search alternatives. Project page: https://itruonghai.github.io/omr.

Modern video generators excel at synthesizing individual clips, but complete video production requires coordinating a long sequence of interdependent creative steps, including scripting, storyboarding, generation, and editing. It further demands persistent asset management and dynamic task orchestration as intermediate outputs, dependencies, and execution states evolve over time. Existing automated systems typically rely on rigid pipelines that are difficult to adapt to diverse inputs and changing workflows, while general-purpose large language models (LLMs) remain unreliable for long-horizon orchestration and multimodal asset routing. We introduce FRAMEWORKERS, a task-centric and workspace-grounded multi-agent framework for open-ended video production. A central Director formulates video creation as dynamic task management, continuously editing a Task Stack to determine which subtask to execute next and which sub-agent to invoke. An Assistant serves as the execution layer, grounding each selected task in a shared Workspace, retrieving the required assets and context, invoking the assigned sub-agent, and persisting the resulting artifacts. Execution capabilities are exposed through modular sub-agents with registered descriptors, allowing new sub-agents to be integrated without redesigning the orchestration workflow. To improve orchestration reliability, we fine-tune the Director via supervised fine-tuning (SFT) followed by Group Relative Policy Optimization (GRPO) for descriptor-conditioned task routing. Experiments show that FRAMEWORKERS outperforms strong LLM planners in routing accuracy, recovers reliably from runtime failures, generalizes to unseen sub-agents without retraining, and achieves higher end-to-end video quality and broader task coverage than fixed pipelines, single-agent systems, and prior multi-agent approaches.

World models let robots imagine possible futures, but exploiting this capability for real-time control is bottlenecked by a representation misalignment: the generative model and the planner operate on decoupled manifolds, so the planner has no shared structure to search over and must instead decode every candidate back into high-dimensional pixel space to evaluate it. This decoding step is a major obstacle to real-time control on physical hardware. In this paper, we present Hydra, a discrete World Action Model that closes this gap by moving the planner, both the sampler and the evaluator, inside the model. Hydra establishes a unified latent manifold over visual states, physical poses, and control actions, then compresses this manifold through modality-specific Vector-Quantized bottlenecks into discrete vocabularies of kinodynamic intents and visual states. Because candidates are now drawn directly from this shared manifold, sampling is informed by the model's own understanding of the observation rather than proposed blind, and evaluation happens natively within the discrete space: candidates are ranked by a Kinematic-Perceptual Cost, without ever decoding to pixels. We term this Discrete Latent Planning (DLP). Because planning over discrete intents alone cannot supply the smooth, continuous commands physical actuation requires, Hydra pairs DLP with conditional Flow Matching, which maps each selected intent to a continuous trajectory for execution. Evaluated on two physical robotic platforms, Hydra outperforms state-of-the-art world models in goal-directed planning, while matching or exceeding the closed-loop execution capabilities of leading reactive foundation policies.

Text-to-Image (T2I) models have recently achieved impressive visual fidelity, yet their evaluation remains constrained by benchmarks that are often difficult to interpret and insufficiently diagnostic. Existing skill-based evaluations tend to overlook critical failure modes that strongly impact usability but fall outside standard taxonomies, such as global incoherence arising from missing parts or physically implausible configurations (e.g., floating objects). In addition, prompt difficulty is typically controlled along a single dimension; either prompt length or the number of elements to generate. To address these limitations, we introduce Imag-Eval, a controlled benchmark designed to assess how T2I models ground compositional natural-language instructions into visual outputs. Unlike prior work that conflates surface linguistic complexity with compositional difficulty, Imag-Eval explicitly seeks to disentangles these factors by independently varying both the number of instances and the combination of constraints (rules), while avoiding error propagation. This design enables fine-grained and interpretable analysis of where cross-modal instruction following fails. Our benchmark comprises 1,140 prompts and 8,842 combined rules, and we evaluate it on several state-of-the-art models. Complementing this analysis with an additional study of over 2,000 prompts from a concurrent benchmark, our results suggest that, for structured skills, compositional difficulty is primarily governed by the number of grounded rules and their binding to instances,, rather than by prompt length alone.

JEPA world models make latent-space planning a practical route to control, but they are built almost exclusively on images. Whether latent prediction survives geometric observations is unclear: point clouds are sparse, unordered, and self-occluded, and with 0.3-15% of scene points moving, the slow-feature optimum of latent prediction compounds with the geometric shortcut of 3D self-supervision. We lift three canonical JEPA designs to point clouds, frozen-encoder, distribution-prior, and action-sensitive, and re-sense the stable-worldmodel benchmark so that only the observation differs from the image baselines. All three plan without collapse: the distribution-prior model is statistically equivalent to its re-evaluated image counterpart on every benchmark, and the action-sensitive model attains the strongest result in our controlled comparison where the most geometry moves. Probing explains why: object positions are almost perfectly linearly decodable and attention falls on the few moving points. Planning withstands heavy dropout never seen in training, though range noise defeats the thinnest scene. Geometry finally makes a commanded 3D target a natural goal interface: we construct the goal latent from the target and the current latent, at no cost in success rate, without a goal observation.

Collecting contact-rich robot experiences at scale remains a major bottleneck for generalizable manipulation. Beyond data quantity, robot learning also requires diverse experiences across embodiments, viewpoints, and scenes. Human egocentric videos provide abundant physical interactions, but each video captures only a narrow slice of experience under a single body, camera trajectory, and environment. We propose AnyWorld, a cross-embodiment world modeling framework that expands a single human interaction into diverse robot-native rollouts without paired human-robot demonstrations. Our model factorizes an interaction into action, camera, and embodiment: action controls capture the motion structure, camera controls specify viewpoint evolution, and the target embodiment context defines the acting body and its interaction geometry. This formulation enables independent recomposition of embodiment, viewpoint, and scene factors, allowing a single model to generate many robot-domain experiences while preserving the underlying dynamics and object interactions. We train the model with large-scale human interaction pretraining followed by mixed-embodiment fine-tuning. Experiments show that our model supports controllable recomposition across embodiments, viewpoints, and scenes, and we further demonstrate that the generated data can improve manipulation performance on the RoboCasa GR1 tabletop benchmark and a real IRON humanoid robot. Beyond aggregate gains, we test whether unpaired human experience can be recomposed into robot-native video-action pairs that target a policy gap. Controlled IRON interventions correct a spurious completion prior and establish language-grounded spatial target selection; an action-only counterfactual intervention fails to learn the latter reliably, showing that both action calibration and visual recomposition are necessary.

Training a video-generation model from scratch is hard for reasons that precede model design. The feedback loop is long: a failure that appears only after a training run can make each attempted fix another run. The data are hard to reach: the corpora and recipes behind strong models are large, heterogeneous, and often unreleased. And scoring is blunt: open-ended generation has no single correct output, and an aggregate score does not by itself establish whether a sample succeeds or which property failed. Dancing Stick Figures is a synthetic video dataset built against these three obstacles. For iteration speed, its 64x64, 64-frame reference task is sized for practical repeated training on a single workstation GPU. For accessibility, the release is a 0.79-GB training tier of 4,020 video clips--1,340 six-second source motions, each rendered from three cameras by a deterministic dataset-generation harness--with checkpoints and a Colab workflow that reruns the reference training pipeline at reduced budget on a 16 GB Tesla T4. For scoring, every frame retains its generating state (ARDY cskel27 joint positions, camera, body parameters, and source motion) and per-pixel depth, surface normals, and part labels. These annotations support dataset-specific metrics for visible topology and part-wise motion; corruptions expose their sensitivities and blind spots.

We present the winning solution to the ACM Multimedia 2026 Grand Challenge on Single-Image Guided Multi-Angle Image Synthesis. It ranks first among 293 registered teams; 56 teams obtained at least one scored submission on the public Phase-A leaderboard. With only 40 training scenes, the challenge requires 26 target views from one RGB model and one forward pass per view; it prohibits explicit geometry, external rendering, chained generation, candidate selection, and post-processing. We identify a critical model-selection failure: shared training and validation scenes make memorization appear as transferable view control. We therefore introduce GoM. Short for Generalization over Memorization, the framework combines scene-disjoint validation, exposure-matched selection, and targeted diffusion adaptation. Its synthesis model adapts a 4B rectified-flow DiT using rank-32 LoRA, optimizer restarts, late-checkpoint averaging, and VAE decoder tuning. More than 300 offline experiments and 24 online submissions show that validation design and training-trajectory control can matter as much as architecture scale in small-data generative modeling.

Natural images concentrate their detail in a small fraction of the frame, yet diffusion models spend a full token on every patch, in every layer and at every timestep. The waste is largest in pixel-space models, with no autoencoder to absorb low-level redundancy first. Probing a pretrained pixel text-to-image transformer, we find its middle-block tokens redundant wherever the image is flat. The redundancy occupies connected, content-shaped regions, and exploiting it requires tokens with the same geometry. Cutting a Hilbert ordering of the patches provides them. Consecutive positions are always image neighbours, so any contiguous run is a connected region whose size and shape follow the content, and grouping in two dimensions becomes a cut in one. Existing reductions each lose part of this. Similarity merging scatters its groups, latent bottlenecks discard position, and skipping deletes what it should summarize. We cut where the model's features change most and pool each run into one region token. Our Region Token Interface (\method{}) adapts a diffusion model to these tokens, with the region count drawn at random during fine-tuning so one checkpoint serves every budget. \method{} leads prior reduction methods at matched budgets, matches dense quality at $2.0\times$ the speed, and stays close at $2.6\times$. The code and models are open-sourced at https://eduardzamfir.github.io/rti

Recent video diffusion models have achieved remarkable generation quality, but high-fidelity results still largely depend on closed-source systems or costly large-scale infrastructure. Test-time scaling (TTS) offers a training-free way to improve lightweight generators by spending additional inference compute, yet existing methods mostly remain within a noise-search paradigm: they sample, select, or perturb denoising trajectories and discard low-scoring candidates after expensive generation. This generate-and-discard process wastes not only computation but also the partial motion, layout, or appearance structure already encoded in recoverable samples. We present GEARS (Guided Editing for Adaptive Recycling Search), a training-free framework that introduces {diagnosis-guided candidate recycling} into video TTS by turning such candidates into editable priors through a generation-evaluation-editing loop. GEARS consists of two collaborative components. The Stage-Aware Scheduler determines what to repair, when to repair it, and which candidates should be preserved, recycled, or discarded. The Candidate Recycler diagnoses recoverable failures from keyframes and multi-dimensional reward feedback, derives candidate-specific repair prompts, and repairs the corresponding candidates through manifold-aware latent SDEdit. The repaired candidates are recycled into the search pool, creating refinement paths beyond standard noise perturbation while preserving useful structure. Under matched NFE budgets, GEARS consistently outperforms existing video TTS methods on VBench, bringing a 1.3B model to a total score comparable to a 14B counterpart, and ablations verify the necessity of adaptive scheduling, diagnosis-conditioned editing, and manifold-aware re-denoising. Code is available on GitHub.

Latent generative models typically follow a two-stage pipeline, training a variational autoencoder for reconstruction and then a generative model on the frozen latent space. Since reconstruction-optimized latents are not necessarily generation-friendly, jointly training both models is an appealing alternative. However, direct end-to-end training remains challenging, as it is prone to latent collapse and faces a generation-reconstruction conflict. We revisit this problem by analyzing how different objectives shape the latent space and identify two key insights. First, the entropy term in the Kullback-Leibler divergence objective is essential for preventing collapse: reconstruction and prior fitting tend to shrink the posterior, while entropy preserves non-degenerate latent uncertainty. Second, reconstruction and generation exhibit asymmetric learning dynamics: reconstruction is fast and strongly supervised, whereas generation is slower and harder to optimize. Based on these insights, we achieve the first direct end-to-end training without latent collapse and propose GenFirst, a simple generation-before-reconstruction strategy. The generative objective first shapes the latent space under weak reconstruction pressure, after which reconstruction is progressively strengthened to recover visual details. We validate GenFirst with continuous autoregressive priors with exact likelihoods and SiT priors with implicit likelihoods. With our end-to-end objective and GenFirst, SiT achieves a gFID of 0.97 with CFG and 1.45 without CFG on ImageNet-256, while MMDiT reaches a GenEval score of 0.90 on text-to-image generation. Beyond image generation, we extend the framework to shared visual latents for generation and representation learning, and to continuous unified text-image generation. These results demonstrate the generality of stable end-to-end latent learning across generative priors and modalities.

While modern generative models excel at modeling complex data, precise inference-time control in conditional generation remains a critical challenge. Classifier-free guidance (CFG) is a primary mechanism for such control, yet it is typically treated as a static tuning parameter. In flow-based models, however, the guidance scale fundamentally dictates the velocity field and the resulting probability path, making guidance selection a dynamic path-optimization problem. We introduce PathGuide, a framework that reformulates scalar CFG selection as an on-policy transport problem. Leveraging the weak form of the continuity equation, we derive a selection criterion with a direct path-correctness interpretation: we prove that if the guided field is weakly equivalent to the exact conditional field along the generated rollout, the sampler's path coincides with the target conditional law. For scalar CFG, this criterion yields a strictly quadratic local objective with an efficient, closed-form selector for each solver interval. PathGuide enables optimal guidance scales to be computed and used online during generation or fitted offline as a reusable piecewise-constant schedule. We validate our method on low-resolution image manifolds and controlled settings across various continuous-time flow constructions, demonstrating that this transport-based selector improves path alignment and sample fidelity over both fixed and state-of-the-art adaptive guidance baselines.

Existing audio-video generation methods produce well-synchronized clips but are limited to short durations, while long-video generation methods extend duration through chunk-based iterative synthesis yet lack audio entirely. Generating long audio-video jointly is fundamentally harder than either task alone: each chunk must simultaneously maintain video temporal coherence, audio temporal coherence, and cross-modal synchronization, whose conditioning signals enter the model through different pathways. In this work, we present Encore for long-form synchronized audio-video generation. Our key insight is to factor this challenge into: (1) local continuity handled by iterative generation with explicit cross-chunk context propagation, and (2) global consistency enforced via reference audio-video signals with shifted position embedding. Building on this design, we propose Adaptive Signal Routing (ASR), which introduces learnable attention biases within self-attention and learnable residual scales on cross-attention outputs, enabling the model to adaptively modulate the influence of each conditioning signal. Trained end-to-end for joint audio-video generation, Encore also supports infinite-length audio-to-video and video-to-audio synthesis at inference by conditioning on the ground-truth modality throughout the denoising process. Experiments on our extended VerseBench for long audio-video evaluation demonstrate that Encore significantly outperforms existing methods in both generation quality and temporal coherence. Code and data for this paper are at https://github.com/shaohua-pan/Encore.

Facial appearance editing powers popular applications like FaceApp and Photoshop. Generative Adversarial Networks (GANs) and 3D Morphable Models (3DMMs) have been widely used for facial editing. GANs can perform varied facial edits (e.g., changing hair color, hairstyle), but often produce unstable edits. 3DMMs produce stable edits, but can only alter pose and facial expression. Recently, text-guided diffusion models like Nano Banana have become popular for image editing. Text-guided models are a compelling alternative to GANs and 3DMMs since they can produce both stable and varied image edits. While text-guided models have been widely tested for whole-scene edits (e.g., ``make the woman play a guitar''), they have not been comprehensively tested for facial editing. We conducted the first large-scale evaluation ($\sim1$M images evaluated) of six popular text-guided models on a sequential facial editing task. We present Face-Edit-Attributes, the largest collection of $169$ facial editing attributes focused on hair, accessories, and pose edits. We compared model performance using two popular celebrity face datasets: CelebA and CelebSET. Our results show that most models performed hair and accessory edits well, but struggled with editing pose. All models over-edit (e.g., changing hair color when asked only to change the hairstyle). We also evaluated demographic biases in each model. Our results show surprising biases in overediting: almost all models created more overedits for dark-skinned male faces and old faces. The code and data for our results (including our repository of $\sim 1$M images) can be accessed \href{https://github.com/rahul1801/Face-Edit-Bench}{\textcolor{blue}{here}}.

Video world models increasingly serve as data engines, action planners, and simulators for embodied AI, but conventional embodied world model (EWM) benchmarks lack a unified 3D-grounded protocol for establishing whether generated rollouts preserve the underlying 3D scene state or translate into executable actions. We introduce RoboPhys-3D, a 3D-grounded EWM benchmark built on RoboTwin 2.0, covering 50 manipulation tasks across four regimes, with 5,000 episodes and 25,000 multi-view ground-truth videos. A defining feature of RoboPhys-3D is that generated and ground-truth videos are processed through the same 3D reconstruction pipeline, enabling reconstruction-induced error to be distinguished from generation-induced error. The RoboPhys-3D benchmark organizes 50 complementary metrics into 18 sub-dimensions across four levels: pixel-level fidelity, 3D geometry consistency, state-level understanding, and task-level completeness. We further introduce Average Full Score, a hierarchical score averaging all 50 metrics for comprehensive evaluation, and RoboPhyscore, a compact task-aligned score averaging the metrics most strongly correlated with task success. Among the four representative video world models, Cosmos 3 achieves the highest RoboPhyscore (0.6330, 92.7% of ground truth), while state- and execution-grounded metrics reveal substantial failures that perceptual and vision-language model-based judgments fail to capture. RoboPhyscore further exhibits strong agreement with human evaluation (Pearson r = 0.9761 and Spearman \r{ho} = 0.8962), demonstrating the importance of grounded, execution-aware evaluation for EWM capability.

While Large language models (LLMs) incorporate user personalization signals to improve usability and helpfulness, they increasingly shift from providing balanced, informative responses toward optimizing for user satisfaction when conditioned on personal context such as conversation history, inferred preferences, and user profiles. Specifically, we identify three emerging risks: (1) irrelevant personalization, where models reference personal information in unnecessary contexts; (2) preference narrowing, where models reinforce informational echo chambers; and (3) sycophantic bias, where models agree excessively with user opinions. As a result, models may reference personal information in contexts where it is unnecessary, inadvertently collapse response diversity, or agree excessively with user opinions. Despite the growing use of personalization in AI assistants, there has been limited systematic evaluation of its potential side effects. To bridge this gap, we propose PRISK, a dynamic evaluation framework with automated data generation and tailored metrics that uncovers systematic limitations in current LLM personalization and how personalized information shapes its responses. Our empirical analysis across 13 LLMs demonstrates the presence of user profiles and retrieved memories consistently exacerbates biases, resulting in an average drop of 45.9% in irrelevant personalization, 41.7% in preference narrowing and 61.7% in sycophantic bias.

Speech and audio LLMs are often evaluated by asking whether a waveform prompt beats an automatic speech recognition (ASR) transcript. For known closed-set tasks, that comparison conflates two factors: access to acoustic evidence and the need to call a generative audio model. We evaluate this distinction as a controlled call-decision problem. For each example, a policy chooses among keeping a transcript label, using encoder evidence from Contrastive Language-Audio Pretraining (CLAP), Audio Spectrogram Transformer (AST), or WavLM, and calling Qwen2-Audio, Qwen2.5-Omni, or MOSS-Audio; the decisive ablation removes all generative actions while keeping the selector and development protocol fixed. On VocalSound, transcripts reach 0.296 accuracy, so waveform information is needed. Yet supervised CLAP and WavLM controls reach 0.850 and 0.854 with no generative audio calls. A selector with generative actions reaches 0.925 accuracy using 12.5% calls, compared with 0.921 for the matched no-call selector (paired difference 0.004; 95% CI [-0.025,0.033]). Agreement and stacking features improve weaker selectors but do not beat the strongest no-call control. For known-task endpoint claims, the relevant quantity is the marginal value of the generative call after transcript and encoder evidence have already been used.

Text-to-image (T2I) models often generate the wrong number of objects, yet existing benchmarks are too small or weakly controlled to explain why. We introduce NumBench, a benchmark of 640{,}000 prompts spanning 1{,}600 categories and counts from 1 to 100. Its factorial design varies object composition, spatial guidance, and appearance conditions while balancing counts and category exposure. We also develop a process model in which requested instances compete for a finite set of resolvable image regions. The model predicts a near-quadratic collision deficit at low occupancy and shows how coordinated placement reduces it. For scalable evaluation, we propose the Confidence-Weighted Numeric Precision Score (\cwnps), which aggregates three calibrated detectors and discounts uncertain proposals. Across five commercial systems, two open models, and two specialized counting methods, performance declines sharply with requested count; all evaluated methods are weak above 50 objects. Count range has the largest measured effect, followed by layout and composition. Grid guidance is strongest among guided layouts, consistent with the coordination prediction, although the analysis does not establish collision as the sole cause. A 14{,}400-image human study supports automated evaluation through count 50, while results on 243 natural-language prompts show transfer beyond NumBench templates.

People read a humanoid robot's motion in social settings not only for the action performed but for the affect conveyed. Motion carrying that affect has so far been generated for human avatars, where style is taken from a reference clip or an emotion word, neither of which can be quantitatively parameterized. We present PAMoR, which turns affect into a measured control parameter: a valence-arousal (V-A) coordinate computed natively on robot kinematics. It is obtained in closed form from postural expansion and movement energy, and these measurements serve directly as generation conditions, with no human annotation. An action prior and two affect priors, trained in a shared latent space, are composed at each denoising step: the action prior fixes what is performed, the affect priors modulate how. Whole-body motion rolls out autoregressively on a 29-DoF Unitree G1 in real time, with action and affect both editable. Generated motion tracks the commanded V-A over its full range while text-to-motion fidelity still matches text-only baselines. In a perceptual study, raters identify the commanded emotion on 0.38 of trials, above both baselines and approaching the 0.44 reported for acted human bodies.

Video re-shooting re-renders a monocular video of a dynamic scene along a user-specified camera trajectory, and the dominant recipe supplies the target geometry explicitly: per-frame depth lifts the source video into a 4D point cloud, which is rasterized along the trajectory into a point cloud render. Because the render and the source video are both handed to the network as visual conditions, they compete at every denoising step, leaving the model with a trust dilemma --- how much of the render to believe --- which can degrade trajectory control or visual quality on data outside the training distribution. We argue that a render already pixel-aligned with the target view does not need to be supplied as an explicit conditioning stream at all. We propose MANIFOLD4D, which injects the render directly into the initial noise of flow matching, so that generation no longer departs from standard Gaussian noise but from a new noise manifold carrying geometric information, leaving the source video as the only visual condition. The render is thus used exactly once, and the network is never asked to learn how to read it; in subsequent denoising steps the model can focus on the source video. On our DAVIS-Traj benchmark and on the Vista4D evaluation set, MANIFOLD4D attains the best camera-control accuracy on every metric, lowering rotation error by 25% and 27% and translation error by up to 32% over the strongest baseline, while matching it in video fidelity and leading on real-world novel-view photometric quality. In a user study, our method achieves clear advantages in trajectory following and dynamic consistency. The gap widens as the yaw amplitude grows past the training range, and the model still recovers correct dynamic motion from the source video when the render is deliberately corrupted, confirming that the geometric prior guides generation without overriding it.

Image-conditioned 3D generation has advanced rapidly, yet existing evaluation protocols largely judge rendered-view plausibility and semantic alignment, overlooking whether a generator forms a stable 3D interpretation and produces assets usable in graphics pipelines. We introduce Cyc3D, a multidimensional benchmark that evaluates image-to-3D generation along two complementary axes: Cross-View Object Consistency and Representation Quality. At the asset level, Cyc3D measures whether object identity remains semantically coherent across rendered viewpoints. At the model level, we propose View-Cycle Structural Consistency, a closed-loop render-regenerate-align protocol that repeatedly re-observes a generated asset from novel views and quantifies geometric, perceptual, and semantic drift across generations. To assess native asset usability beyond rendered appearance, Cyc3D further evaluates geometric structure, reference-image fidelity, mesh discretization and efficiency, and UV parameterization quality. Together, these diagnostics expose failures obscured by a single perceptual score and provide interpretable evidence of both model instability and representation defects. Experiments on five representative image-to-3D systems show that closed-source feed-forward models consistently outperform open-source optimization-based baselines in geometric fidelity, mesh quality, and cycle stability. Nevertheless, even the strongest methods achieve cycle-stability scores below 48, revealing a persistent gap between visually plausible generation and robust 3D object understanding.

Predicting robot videos requires both precise motion reasoning and preservation of high-frequency appearance, yet monolithic pixel models entangle these objectives and often conceal their progress behind a strong last-frame baseline. We present AcrossVAM1.0, a lightweight, text-assisted video action model that factorizes future prediction into object-centric motion and dense appearance. A frozen SAM3-DLP codec decomposes four context frames into semantic particles for the robot, arm, and gripper, together with a background latent. A 0.28M-parameter spatio-temporal Transformer aligns particle identities, rolls their states forward, and is modulated by a frozen OpenCLIP instruction embedding through FiLM. A causal dual-stream decoder combines particle-rendered motion with appearance encoded exclusively from the last observed frame; a residual refiner and learned delivery mask produce five future frames without access to future appearance. On our VRS benchmark constructed from diverse real-robot trajectories, particle dynamics reduce trajectory error by 21.0% over persistence. Across three delivery-mask seeds, AcrossVAM1.0 improves future-frame PSNR/SSIM from 19.97/0.796 to 20.573/0.8004, while raw particle generation improves motion-region PSNR from 11.89 to 13.23. The delivered model does not yet beat persistence in LPIPS, and correct-versus- shuffled language changes trajectory error by only 2.8--3.1%. We report these limitations alongside oracle, negative-control, multi-seed, and per-robot analyses. The results show that explicit particle dynamics are a promising low-dimensional interface for robot video prediction, while robust language grounding and appearance delivery remain the principal open challenges.

Cross-identity character animation aims to drive a target identity from a reference image to follow the motion of a source character from a driving video. The core challenge lies in the inherent entanglement of two capabilities: cross-identity spatial mapping (aligning position, scale, and skeletal proportions) and motion control (refining joint articulation, volumetric consistency, and view coherence). We introduce Reference-Aware Structural Alignment (RASA), a framework that disentangles spatial mapping from motion control by injecting structured priors into a Diffusion Transformer (DiT). Our approach has two stages. First, a Spatial Prior Calibrator (SPC) fuses reference identity with driving pose to generate a spatially grounded initial noise latent, ensuring correct positioning, scaling, and alignment with the driving skeleton. Second, an Inherent Motional Guider (IMG) encodes shape-agnostic SMPL articulation parameters into a semantic motion vector beyond appearance-biased 2D keypoints. Injected into intermediate DiT layers, this vector complements the base pose condition for anatomically consistent articulation and view-aware volumetric refinement. We curate CIM-Bench, a high-quality benchmark with rigorous curation, for evaluation. Extensive experiments show RASA significantly outperforms state-of-the-art methods in motion fidelity and visual quality. Our work establishes a new paradigm showing disentangled spatial and motional priors are key to robust character animation. Project page: https://hidream.ai.github.io/RASA/

Autoregressive text-to-image generation has recently achieved remarkable progress, offering high-fidelity synthesis via a unified generative framework. However, fine-grained semantic control remains challenging due to the attribute entanglement and the misalignment between textual and fine-grained visual representations. In this paper, we introduce Attribute Token Arithmetic (ATA), a method that enables disentangled and continuous attribute control in visual autoregressive modelling. Inspired by the vector arithmetic property observed in word embeddings, ATA identifies semantic directions corresponding to visual attributes (e.g., aging, fatness, emotion) directly within the pretrained autoregressive latent space. These directions are learned from a single reference image, without model retraining or large-scale supervision. During generation, attributes can be continuously adjusted and compositionally combined through simple arithmetic operations with other attribute tokens. Extensive experiments demonstrate that ATA achieves identity-preserving, fine-grained, and multi-attribute adjustment, outperforming existing autoregressive editing baselines in controllability, generality, and computational efficiency. Our code will be available at https://github.com/Madaoer/ATA.

Recent generative approaches to geometry estimation adapt pretrained image diffusion models and treat the task as image-conditioned generation. Leveraging off-the-shelf image diffusion models, they either (i) train task-specific geometry models (for depth and surface normal estimation) independently, losing the opportunity of exploring the intrinsic correlation of these geometric targets, or (ii) jointly fine-tune modified image diffusion backbones (e.g., altered self-attention), which typically demands substantial labeled data. To overcome these limitations in a principled fashion, we repurpose pretrained video generative models as a unified and data-efficient framework for geometry estimation, formulated innovatively as a next-frames prediction task. Our method, GeoNeXt, inherits naturally structured knowledge and richer priors from the video model, while further adapting them for joint modeling of images and geometry targets (image <-> geometry), enabling more data efficient and effective learning of geometry. Extensive experiments validate our method for zero-shot monocular depth and surface normal estimation across diverse datasets, outperforming both previous task-specific and unified generative competitors while using substantially less training data. Notably, our method rivals discriminative state-of-the-art approaches trained on over 100x more data and even standouts on several benchmarks.

Autoregressive video diffusion enables scalable long-video generation by producing chunks from a bounded recent context. While recency-based caching preserves local continuity, it evicts historical cues needed when subjects, objects, scenes, or attributes reappear. Existing memory mechanisms expose models to nonlocal history, but access alone does not ensure effective use. Our analysis reveals that video DiT layers exhibit distinct preferences for current, recent, and distant context, suggesting that long-range memory requires deciding both what to retrieve and where to use it. We introduce LayerRecall, a current-conditioned, layer-selective memory router that retrieves relevant historical K/V states and injects them only into backbone-specific memory-sensitive layers while preserving local attention elsewhere. To reduce reliance on scarce high-quality long-horizon videos and explicit memory-allocation labels, we further propose Cross-Horizon Prediction Matching (CHPM), which uses a privileged long-context reference to supervise the bounded-memory router in prediction space. Across 100 multi-shot evaluation prompts, LayerRecall achieves the best overall results on MemoBench and MovieBench while matching its backbone on VBench-Long, demonstrating stronger long-range recovery without sacrificing local continuity. Qualitative analyses further reveal memory-guided self-correction, whereby initially mismatched local attributes return to their historical appearance without resetting ongoing motion or scene structure. Additional analyses show cross-backbone portability and negligible inference overhead.

Autoregressive video diffusion models enable streaming generation through sliding-window attention, but each generated block is conditioned on previously generated content, causing appearance and motion errors to propagate recursively over time. Historical key-value (KV) memory preserves earlier subject and scene states and helps maintain long-horizon consistency. However, retaining every generated state creates a historical archive that grows continuously with the rollout, while recurrent states repeatedly add redundant coverage. To address this problem, we propose DensityKV, a training-free historical KV bank management strategy. DensityKV maintains a separate token-level KV bank for each attention head and measures local redundancy among the post-RoPE keys that directly parameterize attention routing using Soft-Riesz density. By constraining neighborhood-density growth after states enter the bank, DensityKV limits repeated historical accumulation while preserving coherent states from each completed generation block. Experiments across three autoregressive video generation backbones and multiple generation lengths show that, at the same upper bound on historical KV capacity, DensityKV improves long-horizon consistency and generation stability while keeping persistent historical storage bounded independently of rollout length.

Enabling humanoid robots to respond to human speech with synchronized and semantically meaningful gestures is fundamental to natural human-robot interaction. However, this task faces three critical barriers: the scarcity of semantically rich datasets, the "modality eclipse" where models ignore audio cues in favor of kinematic inertia, and the sim-to-real gap regarding physical safety. We propose RoboGesture, a robot-centric framework that co-designs data, modeling, and control to power a complete interactive human-humanoid system in which the robot listens, responds, and gestures in real time. We first establish the RoboGesture dataset featuring over 300 gesture categories and develop an automated pipeline to synthesize large-scale collision-free, robot-specific audio-motion pairs. Our architecture features a Hierarchical Semantic-Acoustic Aligner that extracts multi-granular prosodic and semantic cues directly from raw audio tokens. These cues drive a Streaming Conditional Motion Generator based on a diffusion transformer with conditional flow matching. To ensure high responsiveness, we introduce Anti-Inertia CFG Masking, which prevents the model from collapsing into repetitive historical patterns by compelling it to proactively mine control signals from the audio modality. Finally, an MPC-based safety filter ensures real-time, collision-free execution on physical hardware. Experiments on a Unitree G1 humanoid demonstrate that RoboGesture generates safer, more rhythmic, and more semantically appropriate responses compared to state-of-the-art baselines.

Long-horizon video generation is evaluated with whole-frame metrics that reward motion and temporal consistency. For fixed-camera nature scenes this creates an ambiguity: motion of water, fire, smoke, or rain is desirable, whereas motion of the background is an error. A system can therefore score well on motion while its scene drifts, or on consistency while its flow stagnates. We introduce SNF-Bench, an evaluation framework for long-horizon fixed-camera generation that partitions each scene into static support and dynamic flow and reports static fidelity, flow persistence with absolute magnitude, and drift leakage separately, never as one score. Drift leakage is interpretive context rather than a headline measurement. Each factor is validated mechanistically rather than by correlation with preference: we inject global translation, rotation, and scale drift and progressive late freezing at known severity into real generations, and require each factor to respond in its stated direction and to remain selective against corruptions it does not target. Auditing publicly released long-horizon text-conditioned checkpoints under one recorded common inference configuration, plus an image-conditioned track with released-pipeline references and a deployment-sensitivity panel, we find that whole-frame motion and static-region drift induce near-opposite orderings of the same outputs. At maximum controlled translation, fBD and NBF rise to $1.86\times$ and $1.32\times$ baseline, but whole-frame Dynamic Degree reaches only $1.07\times$---rewarding the corruption. SNF-Bench measures where motion occurs and whether it persists; it does not measure physical realism. Project page: https://minar09.github.io/snfbench/.

GUI agents trained with reinforcement learning (RL) have showcased strong environment learning capabilities on mobile platforms. However, RL typically demands extensive real-environment interactions, leading to high resource costs and instability, especially in GUI scenarios. To address these, we propose WM-R1, the first reinforcement learning framework that trains mobile GUI agents with world models instead of real environments. Specifically, world models serve as the source of state transitions during all rollouts, replacing the real Android environment within the training loop. WM-R1 also embeds world models directly into the thinking process, enabling agents to reason about the consequences of candidate actions before committing to the final action. Crucially, WM-R1 eliminates the need for real-environment interaction, supports massively parallelized and step-level granularized trajectory generation grounded in world models, and introduces a multi-dimensional rule-based reward that jointly optimizes task success, trajectory efficiency, and world model utilization. For efficient training, we curate a high-quality dataset of 2000 challenging tasks. Experiments on Android mobile benchmarks demonstrate that WM-R1-trained agents significantly outperform GRPO-only baselines and inference-time simulation methods. Code is available at https://github.com/genalyu/WM-R1 .