论文库

Frontier video generation models work in a VAE latent space that compresses video in both space and time. This compression made training tractable, but now limits high-resolution, controllable, and interactive generation -- doubling resolution quadruples token count, conditions apply to groups of frames rather than individual ones, and each response must wait for the rest of its multi-frame block. We present PixVid, a simple yet effective method that lifts a pretrained latent video model to directly generate every frame in pixel space. PixVid keeps the diffusion transformer and replaces the VAE decoder with a convolutional head that renders its outputs into pixels. We find that predicting clean video rather than velocity, together with a larger noise shift, is key to quality. Finetuned from Wan2.1-14B, PixVid is competitive with the latent baseline within $\sim 1.3$M training clips. Working framewise in pixels, it generates at $2\times$ higher resolution with the same per-frame token count, keeps depth-conditioned video editing aligned under fast motion, and responds to each action at the next frame. Project page: https://chenshuo20.github.io/PixVid/

How multimodal language models should encode visual input remains up for debate. Most models use a pretrained vision encoder, while some recent recipes train the encoder from scratch with the language model or remove it entirely. Vision encoders bundle together three design choices: a visual prior from upstream pretraining, dedicated vision capacity, and pre-fusion placement of that capacity (i.e., before the multimodal backbone). Because published models also differ in data, architecture, and compute, comparing them cannot isolate the effect of each choice. We disentangle these choices via a controlled framework where we pretrain all models from scratch on common data and compare them on visual question answering and language benchmarks at matched lifetime compute. We find that the visual prior is not necessary; an encoder trained jointly from scratch is on par with a pretrained one, without a separate training stage. What matters instead is dedicated vision capacity: an encoder-free model trails on visual understanding even when its backbone is enlarged by the encoder's parameter count. Yet that capacity need not precede multimodal fusion, as vision-only modules added in-fusion to early backbone layers perform comparably to a pre-fusion encoder and can lower time to first token. These results also hold across backbone sizes, when encoder pretraining is treated as free, and when starting from a pretrained language model. Overall, our findings favor dedicated vision capacity trained jointly with the backbone and placed early within it.

Decoder-only Large View Synthesis Models (LVSMs) with KV-cache have recently achieved state-of-the-art quality by regressing novel views from neural networks without reconstructing 3D geometry, while Test-Time Training (TTT) layers save computation but sacrifice quality. We propose Test-Time Baking (TTB), a strategy that learns to bake global scene information from cross-view attention into a lightweight MLP, enabling fast and high-fidelity novel-view decoding. The first key is to repurpose the fast-weight MLP of TTT to learn the cross-view attention mapping explicitly from its input-output pairs, rather than replacing attention entirely, achieving the same rendering efficiency as TTT while delivering substantially higher quality. The second key addresses a modality misalignment between context prefilling and novel-view rendering: we disentangle context tokens into camera-only context rendering tokens to serve as TTB input, and context ground-truth tokens with image content to serve as the TTB mapping target, ensuring the baked MLPs are relieved of the modality generalization challenge between context and novel-view processing. Our method achieves up to **23x** rendering speedup with quality on par with the state-of-the-art across several benchmarks. Project page: https://test-time-baking.github.io

Vision language navigation (VLN) is very challenging for Unmanned Aerial Vehicles (UAVs) since small errors in 3D motion prediction can lead to navigation errors and collisions. Unable to fathom execution outcomes in 3D environments, existing methods often struggle to predict accurate navigation waypoints from language instructions. To solve this, we introduce the Prototypical Visual Imagination framework, shortened as ProtoVis-Nav, which forces the model to imagine an outcome-associated local visual representation for the intended next UAV movement. Specifically, we predict feature prototypes and use their linear combination to estimate the visual feature of the target region. In this manner, the VLN model first imagines the local visual representation of the next intended location, and then uses predicted prototypes as context for generating the final action, promoting alignment between visual context and waypoint outputs. Extensive experiments demonstrate the effectiveness of the ProtoVis-Nav framework and show substantial improvements over existing baselines.

Video diffusion models are able to generate samples with high-quality appearance and textural qualities. However, content produced by these methods is known to exhibit significant physical errors. This is because, in a classical video generation setting, all physical reasoning must take place simultaneously and exclusively within the diffusion process.We propose to resolve these issues by redistributing the task of physical reasoning onto a language model-backed keyframe generation module, which allows for explicit reasoning in language, and fine-grained control over the appearance of the video at important moments in the video. We then condition a video generation model with these keyframes, reducing the role of the diffusion model to the simpler task of interpolating between physically-correct intermediate states, rather than inferring an entire sequence with diffusion. Results demonstrate that our approach significantly outperforms existing approaches, achieving state-of-the-art performance on the Physics-IQ Verified benchmark and PhyGenBench.

Given a set of LEGO parts and an image of an assembled object, our task is to infer the pose of each part to assemble the object. In contrast with traditional 3D reconstruction, doing so requires satisfying discrete physical constraints, which provide structure but render the problem combinatorial and highly sensitive to small pose errors. To learn a prior over how parts fit together, we train a set-conditioned SE(3) flow model on procedurally generated assemblies. We then finetune this model to inject image conditioning, using 1% as much data. However, while the learned flow captures global structure, it does not enforce connectivity. To address this, we introduce an analytic connectivity-field guidance to pull predictions toward connected assemblies at test time. Through this, we demonstrate improved build quality with model size, pretraining, and guidance. We make our models and data available for research purposes at https://kulits.github.io/BrickFlow/

Spatial intelligence requires connecting visual observations to the geometry, objects, and persistent state of an environment. Images alone leave these relationships implicit, while disconnected annotations make it difficult to study appearance, camera motion, and scene changes under a common reference. We introduce Spatial Foundry, a collection of 5,280 furnished residential 3D scene configurations with editable assets and aligned visual, geometric, and temporal data. Its modalities include floor plans, perspective and panoramic RGB, metric depth, point clouds, meshes, instance-level 3D bounding boxes, and camera trajectories paired with video. Shared scene coordinates link observations across views and time; editable source scenes enable additional camera paths and controlled object changes. We describe the data representation and construction workflow, and discuss applications in floor-plan-conditioned 3D generation, geometric reconstruction, instance-level perception, camera-conditioned video generation, shared world modeling, and scene editing. Controlled camera and scene variations support spatial perception and world-model research. The public project page provides videos and an interactive colored point-cloud preview.

Local Binary Patterns (LBP) remain among the most widely used texture descriptors, from industrial inspection to medical imaging. Yet their radius, sampling angles, color transform and thresholds are fixed by hand for each dataset. We present FLEX-LBP, which learns these parameters by backpropagation while every forward pass computes exact binary codes and hard histogram counts. A straight-through estimator acts only in the backward pass, so the deployed descriptor is an interpretable hard-binary LBP with 90 learnable parameters and unchanged inference cost. In a controlled ablation on six color texture benchmarks, the component worth learning is dataset-dependent. One configuration, applied unchanged everywhere, learns the geometry and thresholds with the color transform frozen. It improves the nearest-neighbor accuracy of the identical fixed descriptor by 3.55 points on average over five seeds, significantly on five of six benchmarks and never significantly worse. With a linear discriminant readout, it gains 0.9 points; the fully learned descriptor gains 2.7 and has the highest mean. FLEX-LBP is more accurate than five classical LBP-family descriptors, PCANet and a neural LBP on five of six benchmarks. Its mean nearest-neighbor accuracy exceeds that of frozen ImageNet ResNet-50 features (23.5 million parameters). On the Kather colorectal cancer histology benchmark, the fully learned descriptor reaches 93.8%, significantly above the best classical descriptor (Completed LBP, 92.8%). The learned radii adapt in directions that recur across seeds. Noise drowns single-pixel comparisons and illuminant change flips inter-channel ones; median sampling or gray-world normalization restores 96-97%. The code is available as a drop-in PyTorch layer at https://github.com/medalimoussa/FLEX-LBP.

Recent advances in diffusion and flow matching models have highlighted a shift in the preferred prediction target -- moving from noise ($\varepsilon$) and velocity (v) to direct data (x) prediction -- particularly in high-dimensional settings. However, a formal explanation of why the optimal target depends on the specific properties of the data remains elusive. In this work, we provide a theoretical framework based on a generalized prediction formulation $u=kx-(1-k)n$, where $x$ represents clean data and $n$ denotes noise. This formulation accommodates arbitrary targets, of which $\varepsilon$-, v-, and x-prediction are special cases. We derive the analytical relationship between data's geometry and the optimal prediction target, offering a rigorous justification for why x-prediction becomes superior when the ambient dimension significantly exceeds the data's intrinsic dimension. Furthermore, our controlled analysis reveals how time conditioning and loss variations systematically modulate this optimal objective. While our theory identifies dimensionality as the governing factor for the optimal prediction target, the intrinsic dimension of manifold-bound data is typically intractable to estimate in practice. To bridge this gap, we propose k-Diff, a framework that employs a data-driven approach to learn the optimal prediction parameter k directly from data, bypassing the need for explicit dimension estimation. Extensive experiments in both latent-space and pixel-space image generation demonstrate that k-Diff consistently outperforms fixed-target baselines across varying architectures and data scales, providing a principled and automated approach to enhancing generative performance.

Interactive video-to-video applications require real-time generation while maintaining long-range temporal consistency. However, recent methods achieve speed by restricting attention to a fixed-length causal window. This limits long-range consistency whenever relevant history falls outside the context window. To address this limitation, we introduce vid2vid-long, a correspondence-based approach grounded in a simple observation. Video-to-video translation provides an explicit cross-frame reference: the regions that match in the input video should match in the output video too. Concretely, we compute patch-level correspondences in the input, guiding retrieval of previously generated outputs in the corresponding regions, and condition the model on these patches through correspondence cross-attention. Our experiments demonstrate that vid2vid-long consistently improves long-range temporal consistency across recent video models while preserving generation quality and real-time performance.

Generative model-based approaches to linear inverse problems typically use the acquired noisy measurement to condition or guide sampling in the image domain. In this work, we introduce Dual Flow, a novel joint-flow model for solving linear inverse problems by jointly estimating the clean image and corresponding noise-free measurement from the observation. Specifically, Dual Flow formulates image reconstruction and measurement denoising as a coupled estimation problem, using coupled flows in the image and measurement domains. The acquired noisy measurement is used only to initialize the measurement state; thereafter, image and measurement states are jointly updated based on their current values. By construction, the final image and measurement estimates are guaranteed to satisfy the underlying forward model. We derive the sampling updates for the joint-flow which involves the conditional mean of the clean image given the joint state. We further derive a closed-form expression for a sufficient statistic of the joint state and show that the required conditional mean can be obtained by solving an image-domain Gaussian denoising problem involving operator-dependent, correlated noise. We train a velocity-prediction network using this formulation to estimate the conditional mean from the derived sufficient statistic. Extensive experiments across various linear inverse problems and image datasets demonstrate state-of-the-art performance in both distortion metrics and perceptual quality. Dual Flow requires no hyperparameter tuning beyond the number of inference steps and achieves shorter reconstruction times than competing methods. Additional analyses assess robustness to model mismatch, as well as the ability to generate diverse plausible reconstructions and characterize reconstruction uncertainty.

Multi-modal three-dimensional (3D) medical imaging data, derived from ultrasound, magnetic resonance imaging (MRI), and computed tomography (CT), provide a widely adopted approach for non-invasive anatomical visualization. However, accurate modeling depends on surface reconstruction and frame-to-frame interpolation, where traditional methods often struggle with image noise and incomplete information between sparse frames. To address these challenges, we present \our{}, a novel framework based on Gaussian Splatting (GS) designed for high-fidelity 3D anatomical reconstruction. Uniquely, \our{} employs a multi-task architecture that simultaneously performs frame interpolation and segmentation using a unified geometric representation. By coupling these tasks, the model leverages dense signals from image synthesis to regularize the geometry, enabling high-quality surface extraction even from a limited number of input frames. Specifically, medical data are modeled as Folded-Gaussians with dual color attributes, supported by an In-Between Frame Regularization (IBFR) mechanism. Experimental results demonstrate that \our{} achieves higher metric scores than implicit neural representations and improves interpolation quality. Repository: https://github.com/gmum/MedGS

Humans naturally parse visual scenes into coherent individual objects rather than scattered features, a capability cognitive science calls object binding. While recent work shows that pretrained Vision Transformers (ViTs) intrinsically acquire this ability, how native self-attention computations mechanistically keep object instances separate remains unresolved. By studying this across eight pretrained ViTs, we found that instance separation is expressed more strongly in attention routing than in token content: selected query-key relations distinguish individual objects more reliably than value similarity or residual representations. Inside heads, patches belonging to one individual tend to read from a common image region, which we term their attention address. Visually distinct parts of one person share an address, whereas identical duplicated instances receive separate, movable addresses. These shared destinations arise in part because query interactions with the spatial structure of keys direct attention toward consistent regions for one object and elsewhere for another. In DINOv2 and most other backbones we tested, editing this query component at the selected head redirects its attention address toward another individual. Grouping patches by this query component separates same-category instances, reaching .615 FG-ARI versus .289 for the same clustering of final-layer features, which tend to merge them.

Most multimodal sentiment analysis algorithms focus on point-wise regression. While straightforward, it uses absolute values to represent subjective sentiment and overlooks the ordinal nature of sentiment, resulting in unstable predictions and poor correlation. Inspired by the way humans perceive sentiment, pair-wise ordinal learning frameworks emerged to address this gap, capturing relative order by learning from comparisons. However, they have two limitations: First, they assign uniform importance to all pairs, failing to adaptively focus on hard-to-rank samples. Second, they employ static ranking margins, which fail to reflect the varying semantic distances between sentiment groups. To address this, we propose a Group-wise Ranking and Calibration Framework (GRCF) that introduces the Adaptive Error-weighting and Dynamic Margin mechanism, which resolves these limitations by simultaneously preserving the relative ordinal structure and ensuring absolute score calibration. Specifically, Stage 1 introduces a Group-aware Ranking Loss to build a fine-grained ordinal structure by assigning higher optimization priority to confusable pairs. Stage 2 then employs an MAE-driven objective to align the prediction magnitudes. To validate its generalizability, we extend GRCF to classification tasks, including Multimodal Humor and Sarcasm Detection. It achieves state-of-the-art performance on regression datasets while also showing strong generalizability in classification tasks.

Hyperspectral images (HSI) possess discriminating spectral characteristics with respect to land utilization due to high-dimensional information. Recently, unsupervised representation learning using graph neural networks has unveiled these utilization-specific complex spectral relations. This article has three main contributions. First, it proposes an index, the Mean Absolute Local Neighborhood Inverse Spectral Index (MALNISI), which measures the similarity between two pixels based on their neighborhood information. The index is used to construct a weighted adjacency matrix, giving the HSI a graph structure. Second, it proposes an upgraded Graph Attention Network that learns compact latent representations of hyperspectral data from the constructed graph. In this network, an enhanced multi-head attention mechanism incorporates the MALNISI weights as an auxiliary feature to emphasize learned attention between neighboring pixels. To alleviate over-smoothing and preserve fine-grained spectral details, the architecture integrates residual connections, skip projections, and feature normalization, which together enhance representation capability, stabilize learning, and accelerate convergence. Third, it proposes a weighted contrastive loss function that incorporates neighborhood-aware MALNISI between connected pixels, encouraging strongly connected pixels in the high-dimensional space to remain close in the latent space, and weakly connected pixels to stay apart. The resulting discriminative low-dimensional embeddings are then clustered using spectral clustering. Experimental analysis on six benchmark HSI datasets demonstrates the efficiency and necessity of the three aforementioned contributions, and the comparative study highlights superior performance over fourteen state-of-the-art approaches.

Distribution matching distillation (DMD) has emerged as a leading paradigm for distilling diffusion models into few-step students, where an auxiliary fake score model tracks the student's distribution and the discrepancy between the teacher and fake predictions drives the student toward the teacher. In multi-step distillation, although only the student's final output is delivered, existing methods train the fake model on intermediate states or across all denoising steps. Consequently, the student's update signal becomes entangled with intermediate distributions, which can distort the matching direction and remain nonzero even after the student's final outputs already match the teacher distribution. To resolve this mismatch, we propose Endpoint Distribution Matching Distillation (eDMD), which trains a single shared fake model exclusively on the student's final outputs and uses it to correct every denoising step. Our theoretical analysis proves that this endpoint reference guarantees that the distribution matching term vanishes if and only if the student's final output distribution matches the teacher's, regardless of intermediate distributions. Beyond the fake reference, we further strengthen teacher supervision through a precise teacher (PT) that replaces single-step guided predictions with short multi-step trajectories on the final noise interval, paired with a calibrated query density for intermediate steps. Experiments on SD3-Medium, FLUX.2-klein (4B and 9B), and Qwen-Image-2.1 demonstrate that eDMD consistently outperforms existing distribution matching methods for 4-step generation.

Reference-frame stabilization registers every frame of a hovering-drone video onto one reference frame, keeping object positions comparable. The registration pipelines it relies on are selected on feature-matching benchmarks in which only the camera moves. Over a busy intersection, vehicles moving together supply mutually consistent correspondences that can draw the robust homography estimate toward their motion: consensus capture, which inlier statistics cannot flag. A two-part benchmark measures it and the object masking meant to prevent it. A static protocol scores ten pipelines under exactly known homographies on 4K aerial frames, tuned on a site-disjoint split at a matched resolution and feature budget. A dynamic protocol scores them at an offline and a real-time budget on hovers selected from the video alone to span the amount and coherence of moving foreground, plus a stationary control. It measures background alignment, object position error and the share of a moving group's displacement an estimate follows, against a pseudo-ground-truth reference audited with manually annotated ground control points, which also score every candidate. On static scenes, tuned AKAZE, RootSIFT, SIFT and BRISK lead, retuning the matcher thresholds raises the mean average accuracy (mAA) by at most 0.0062, and masking costs less than 0.005. Unmasked ORB is captured on 28 of 304 real-time dynamic pairs and on 6 offline; its real-time captured estimates have a median corner error of 18.8 px yet typical inlier counts and ratios. Masking removes capture for every pipeline and leaves the leading pipelines unchanged. Code, data and results are released.

Within the CG animation production pipeline, the challenges artists tackle in Lighting can be immense. Even minor adjustments require re-rendering massive scenes with slow offline renderers; global illumination has to be sampled and complex shaders have to be evaluated, leading to iteration times of minutes to hours per frame. To accelerate this process, we introduce a novel neural render proxy (NRP) that enables differentiable relighting of static scenes with fixed camera and materials at interactive rates. Our main insight is the decoupling of traditional rendering into path sampling and emission computation. From a single, light-agnostic render pass, we collect light transport data in the form of path samples. This enables rapidly sampling lighting conditions on the fly, and training a scene-specific lightweight neural network that learns how light is transported from any scene location to any image pixel. This approach is compatible with non-differentiable production renderers, induces minimal memory requirements during inference, and scales only with resolution and the number of light sources and parameters, but independently of scene and appearance complexity. Our evaluation demonstrates interactive frame rates for relighting (~30-60 Hz) while closely approximating the visual fidelity of ground-truth path tracing. In addition, the differentiable NRP enables fast, gradient-based inverse workflows, allowing artists to efficiently solve for lighting parameters from intuitive image-space edits or generative targets.

The quadratic complexity of the attention mechanism severely limits the context scalability of Video Diffusion Transformers (DiTs). We find that the highly sparse spatio-temporal attention patterns exhibited in Video DiTs can be naturally represented by the Monarch matrix. It is a class of structured matrices with flexible sparsity, enabling sub-quadratic attention via an alternating minimization algorithm. Accordingly, we propose VMonarch, a novel attention mechanism for Video DiTs that enables efficient computation over the dynamic sparse patterns with structured Monarch matrices. First, we adapt spatio-temporal Monarch factorization to explicitly capture the intra-frame and inter-frame correlations of the video data. Second, we introduce a recomputation strategy to mitigate artifacts arising from instabilities during alternating minimization of Monarch matrices. Third, we propose a novel online entropy algorithm fused into FlashAttention, enabling fast Monarch matrix updates for long sequences. Extensive experiments demonstrate that VMonarch achieves comparable or superior generation quality to full attention on VBench after minimal tuning. It overcomes the attention bottleneck in Video DiTs, reduces attention FLOPs by a factor of 17.5, and achieves a speedup of over 5x in attention computation for long videos, surpassing state-of-the-art sparse attention methods at 90% sparsity. Code is available at https://github.com/EricLina/VMonarch.

Differential morphing attack detection (D-MAD) compares a trusted live capture with a document image that may contain a facial morph. We investigate whether labelled validation data can improve a frozen-parameter vision-language evaluator through prompt selection rather than model fine-tuning. Version~1 generates 50 candidate prompts and selects the top-ranked instruction using a security-oriented loss that prioritises missed morph attacks (MACER) while accounting for bona fide classification errors (BSCER), response failures, parse failures, and decision inconsistency. Version~2 recursively revises prompts using aggregate validation feedback. On a subject-disjoint test set of 160 pairs spanning LMA, MIPGAN-2, and PIPE, the selected Version~1 prompt achieves a (19.17\%) morph attack classification error rate (MACER), with (0.00\%) observed bona fide sample classification error rate (BSCER), compared with (70.00\%) MACER for a controlled manual prompt. The paired (50.83) percentage-point difference has a (95\%) subject-bootstrap interval of ([43.33,58.33]). Recursive revision does not surpass the best Generation~0 validation loss; the reported revised descendant obtains (31.67\%) test MACER. All 23 attacks missed by Version~1 are landmark-based (LMA) morphs. Overall, the results show that prompt formulation can substantially influence frozen-parameter VLM-based D-MAD performance, while also highlighting the challenge of recursive prompt refinement under a compact 10-subject validation setting.

Remote sensing change detection usually relies on large collections of temporally aligned, pixel-level annotations, which limits its scalability across regions, imaging platforms and application domains. Recent zero-shot approaches reduce the need for supervision, but they often localize fine-grained changes poorly because their features are coarse or inconsistent across time. We present SpotDiff, a training-free, plug-and-play framework for zero-shot change detection built by composing complementary vision foundation models. SpotDiff uses DINOv3 for semantically robust features, AnyUp to reconstruct them at pixel resolution, and SAM 3 for text-guided region-of-interest proposals, which an inference-time multi-scale strategy (SWIMS) makes reliable on small objects. Change is then measured by comparing the reconstructed features of the two dates inside the proposed regions, with no learnable components, fine-tuning or change labels. To our knowledge, SpotDiff is the first change detection method to use a universal feature upsampler, which alone adds 13.8 F1 over bilinear upsampling of the same features. On LEVIR-CD, SpotDiff reaches 82.46 F1, 2.06 points above the strongest prior training-free method, and on the multi-class SECOND benchmark it improves building change detection by 10.3 IoU and the six-class average by 2.2 IoU. Qualitative results on PlanetScope (3 m) and Sentinel-2 (10 m) imagery show that the same pipeline transfers across resolutions without modification. Our results indicate that composing modern foundation models is an effective route to scalable, label-free change detection.

Language-guided UAV navigation has largely focused on long-horizon route planning, leaving short-range, object-relative flight maneuvers underexplored. To address this gap, we formulate Flying-on-a-Word (Flow), a task requiring a UAV to convert concise natural-language instruction, egocentric visual observation, and current state dynamics into continuous flight actions. Flow encompasses both primitive motion commands and complex object-interactive behaviors, assessing motion intent comprehension and spatial context grounding. To enable policy learning and closed-loop evaluation, we introduce UAV-Flow, a comprehensive dataset comprising 30,692 expert-piloted real-world trajectories across eight distinct motion categories, alongside 10,109 simulation trajectories in the UAV-Flow-Sim environment equipped with task-specific completion criteria. Furthermore, we develop FlowVLA, which undergoes imitation pretraining on mixed real-sim demonstrations and online post-training via Tail-Aware GRPO (TGRPO). By leveraging group reward distributions, TGRPO effectively reinforces high-reward action sequences. Evaluated on 273 held-out UAV-Flow-Sim episodes, pretrained FlowVLA achieves a mean success rate of 75.58%, outperforming the strongest baseline by 6.72 percentage points, which TGRPO post-training further elevates to 93.16% with superior sample efficiency. Physical UAV experiments demonstrate robust execution of object-relative commands in both indoor and outdoor real-world scenes. Together, Flow establishes a novel benchmark and learning framework for language-conditioned flight maneuvers.

The proliferation of user-generated video content has made accurate and robust no-reference video quality assessment (NR-VQA) a highly important challenge. Existing research increasingly relies on architecturally complex multi-branch pipelines or computationally intensive fine-tuning of single-branch backbones. In this paper, we investigate whether a single frozen video foundation model can serve as an effective NR-VQA backbone, and propose HiVE-VQA, the first framework (to our knowledge) to apply a JEPA-based video encoder to this task. HiVE-VQA extracts token grids from multiple depths of a pre-trained V-JEPA 2.1 encoder and stacks them into a hierarchical spatio-temporal volume, which a compact 3D convolutional head fuses across its layer, spatial, and temporal axes, combining low-level distortion cues from shallow layers with the semantic and temporal context captured at greater depths. The encoder remains frozen throughout, so only the head is trained: 5.2M parameters (5.7\% of the total), with no auxiliary multi-modal models and an inference cost that is invariant to video resolution. Trained solely on LSVQ and evaluated without any fine-tuning, HiVE-VQA outperforms 17 state-of-the-art (SOTA) methods in SROCC and PLCC on both intra-dataset (LSVQ-test, LSVQ-1080p) and cross-dataset (KoNViD-1k, LIVE-VQC) benchmarks. These results indicate that hierarchical, depth-wise fusion of features from a single frozen video foundation model provides a more effective and efficient basis for NR-VQA than task-specific multi-network fusion. Code and model checkpoints are available at https://github.com/RobHam99/HiVE-VQA.