video streaming engineers

Abstract: Super-resolution networks are trained for one magnification factor, so several output resolutions require several models. We propose PMS-SR, a progressive single-input multiple-output network that produces x2, x4, and x8 results in one forward pass, with a Transformer backbone on the low-resolution grid. In Stage 1, each scale is predicted as a residual over the detached bicubic upsampling of the previous scale's prediction. With Dirac and zero initialization, training starts exactly from the bicubic cascade, and each head adds only the band the previous scale lacks. In Stage 2, inter-scale mutual refinement couples the scales in both directions through space-to-depth feedback, deformable alignment, and per-pixel fusion, so x8 features refine the x2 output. Trained on DF2K, the 7.95M-parameter model achieves a higher PSNR than every compared multi-scale method at x4 and x8 on all five standard benchmarks. At x8, it outperforms the x8-specific DRN-L on all of them, leads all compared methods on Urban100, and is within 0.13 dB of the best result elsewhere. Against three single-scale SwinIR-M models, PMS-SR uses 22% of the parameters, runs 2.9x faster on BSD100 inputs, and its outputs agree 6.7-12.4 dB more closely. Code and pretrained models will be released at https://github.com/julian135707/PMS-SR upon acceptance.

Abstract: Semantic IDs (SIDs) have emerged as a promising approach for discriminative recommendation by encoding continuous item rep- resentations into compact discrete identifiers. However, existing SID-based methods face two fundamental limitations: semantic mis- match between content semantics and collaborative preferences, and objective mismatch between SID construction and downstream recommendation optimization. To address these challenges, we propose SIDRec, an end-to-end recommendation-aware SID learn- ing framework that jointly models content and collaborative in- formation while aligning SID construction with recommendation objectives. Specifically, SIDRec introduces a self-updating fusion mechanism that explicitly integrates pretrained content represen- tations with collaborative embeddings initialized from a trained recommendation model, enabling preference-aware semantic mod- eling. Furthermore, SIDRec employs an MMoE-inspired joint opti- mization mechanism with shared experts and task-specific gates, allowing recommendation supervision to directly guide differen- tiable quantization and produce task-adaptive discrete representa- tions. Extensive experiments on three public benchmark datasets demonstrate that SIDRec consistently outperforms all competing methods across both AUC and GAUC, achieving up to 5.15% relative improvements over the strongest baselines. Further ablation stud- ies validate the effectiveness and complementarity of collaborative semantic modeling and end-to-end SID optimization.

Abstract: Session-based recommendation (SBR) infers user interest from limited interaction sequences, yet current models show a consistent performance drop when longer histories are available. This degradation is associated with decoder designs that overemphasize recent items while underutilizing distant, yet semantically relevant, interactions, leading to poor modeling of long-range dependencies and multi-interest behaviors. To address this challenge, we introduce MIDA (Multi-Interest Decoding with Adaptive Aggregation for Session-based Recommendation), a lightweight and encoder-agnostic framework that provides a reformulation of the decoding stage. The central design principle is a two-space decoupling: clustering operates on raw item embeddings so that semantic groupings are not distorted by encoder-learned relational biases; encoder outputs are reserved for aggregation, where sequential context remains intact. Concretely, MIDA (i) employs differentiable multi-interest clustering on raw embeddings, preserving semantic structure that encoder representations would otherwise obscure, and (ii) fuses stability and plasticity branches through a gate conditioned on average cluster size, so that the inferred clustering structure directly governs aggregation behavior through a structural coupling across the two stages. Extensive experiments on multiple benchmark datasets demonstrate that MIDA consistently outperforms state-of-the-art methods, with meaningful accuracy improvements while remaining computationally practical.

Abstract: Real-time decoding of omnidirectional images is essential for immersive virtual reality (VR) experiences. Compared with conventional transform-based learned image compression (LIC), overfitted codecs offer lower decoding complexity by representing each image with a lightweight implicit neural representation (INR). However, existing approaches typically operate on the equirectangular projection (ERP), which introduces nonuniform spherical sampling and spatial distortion, leading to redundant sampling near the poles and latitude-dependent receptive fields. To address these limitations, we propose a spherical overfitted codec based on a multi-scale Hierarchical Equal Area isoLatitude Pixelization (HEALPix) representation, which enables equal-area sampling and efficient hierarchical modeling of spherical signals. Building on this representation, we develop a Spherical Inter-Scale Conditional entropy model that exploits spatial dependencies and cross-scale correlations among HEALPix latents to improve coding efficiency. We further introduce a Multi-Ring Spherical Convolution that aggregates neighboring pixels over a nearly uniform spherical receptive field across latitudes. Experiments on a public omnidirectional image dataset demonstrate BD-rate reductions of 16.3%, 26.5%, and 19.5% over VVC in terms of V-PSNR, V-MS-SSIM, and V-LPIPS, respectively. Our method also achieves a 13.1% BD-rate reduction over ERP-based Cool-Chic 5.0 in terms of V-PSNR, while providing 2.6-3.6x faster decoding than the evaluated learned autoencoder codecs.

Abstract: Neural image codecs face a fundamental trade-off between pixel fidelity and perceptual quality. Distortion-optimized codecs preserve pixel-wise fidelity but produce blurred reconstructions at low bitrates, whereas generative codecs leverage learned priors to improve perceptual quality but tend to exhibit saturating PSNR at higher bitrates and require separate models for different bitrate targets. We propose PerQ, which reframes perceptual reconstruction as inverse generative modeling with bounded support. Specifically, we define the quantization cell as the set of latents that yield the same transmitted symbols and derive a transport bound showing that reconstructions from this cell remain within a rate-determined radius of the codec output. This motivates performing generative reconstruction within the codec-induced quantization cell, rather than sampling from an unconstrained image prior. PerQ implements this principle by attaching a rate-conditioned flow-matching compensator to a frozen, multi-rate distortion-optimized backbone, enabling perceptual enhancement while preserving the backbone's coding performance. Experiments on Kodak and CLIC2020 show that PerQ achieves perceptual performance comparable to generative codecs such as MS-ILLM in LPIPS and FID, while maintaining competitive PSNR relative to distortion-optimized codecs such as ELIC. A single model checkpoint supports a wide bitrate range with fast decoding, and the compensator provides consistent gains across three different backbones. The implementation is available at https://jasminepp.github.io/PerQ/.

Abstract: Traditional neural video compression (NVC) methods predominantly follow strictly causal predictive-coding paradigm, treating zero-frame delay as a fixed constraint rather than delay as an explicit design parameter. From an information-theoretic perspective, allowing limited frame delay enables a codec to exploit correlations with future frames within the delay window, thereby reducing uncertainty that is inaccessible to strictly causal prediction and potentially improving the rate-distortion tradeoff. In this work, we formalize this intuition through a rigorous rate-distortion analysis builds on the clique factorization of Markov networks, and propose Graph-Factored Sequential Coding (GFSC), a principled framework for sequential video coding under limited delay. To our knowledge, we provide the first theoretical result showing that, for general $K$-order Markov sources with positive distributions, a $K$-frame delay is sufficient to attain joint-coding rate-distortion optimality over a nontrivial distortion region, a guarantee unavailable to zero-delay causal coding in general. In practice, we instantiate GFSC through lightweight neural modules, allowing existing causal predictive codecs to be augmented without modifying their core architectures. Experiments across multiple NVC models and datasets show that GFSC consistently improves rate-distortion performance over the corresponding causal counterparts. Under matched delay budgets, GFSC also achieves competitive performance compared with representative bidirectional codecs.

Abstract: For decades, conventional video streaming systems have relied on packet-level forward error correction (FEC) codes and retransmissions to ensure reliable video delivery. However, in real-time communications, FEC is inherently inefficient under unknown and time-varying packet-loss conditions, while retransmission can incur excessive latency and fail to meet stringent playout deadlines. Emerging deep joint source-channel coding (DeepJSCC) provides an alternative by learning channel-noise-resilient representations, yet its tight coupling with the physical layer makes it difficult to integrate into today's layered network architectures and protocols. In this paper, we propose TokJSNC, a simple-but-effective tokenized joint source-network coding paradigm for packet-loss-resilient video streaming. Rather than first compressing video into a compact bitstream and then protecting it with packet-level FEC, TokJSNC directly encodes video into variable-length discrete token-stream, whose transmission rate is jointly adapted to video content complexity and network conditions. By exploiting the spatiotemporal correlations among learned video tokens, TokJSNC enables robust video reconstruction over a wide range of packet-loss rates without explicit FEC or retransmissions. Consequently, TokJSNC brings the JSCC-like graceful degradation property to the network layer while remaining compatible with existing network architectures and transport protocols. Experimental results show that TokJSNC maintains an average playback frame rate above 22 fps even at a 70% packet-loss rate, whereas the best-performing baseline drops below 12 fps. Overall, TokJSNC establishes a practical pathway for repurposing modern generative video tokenizers for loss-resilient streaming, while providing a fundamental framework for video token communications.

Abstract: Interactive video generation commonly relies on small bidirectional chunks to reduce response latency. Yet our preliminary experiments show that larger chunks improve video quality and better utilize GPU parallelism, allowing more latents to be sampled with only a modest increase in sampling time. Motivated by these observations, we propose Speculative Sampling, a strategy that combines the modeling power of large bidirectional windows with the low-latency interaction of small ones. It expands the modeling window by appending noise latents as padding and caches the resulting predictions for reuse when the condition remains unchanged. Predictions are delivered incrementally, while the uncommitted future remains replaceable. Our latency analysis shows that this removes the waiting-time lower bound imposed by chunk boundaries, shifting the optimization focus from reducing chunk size to accelerating per-chunk sampling. Bidirectional window size thus becomes a decision variable for jointly optimizing visual quality and sampling time, rather than a necessary constraint on low-latency interaction. This opens a path to serving large bidirectional models in interactive streaming systems without complex streaming post-training.

Abstract: Training-free diffusion-based compression methods repurpose the generative priors of a diffusion or a flow model to compress data without retraining them as task-specific codecs. Recent codebook-based approaches such as Turbo-DDCM achieve strong perceptual quality at ultra-low bitrates, but their reliance on dense Gaussian codebooks remains costly in both memory and computation, and scales poorly with the large latent dimensions of modern image and video generators. Their combinatorial bitstream encoding further becomes computationally heavy at this scale. We introduce H-DDCM, a scalable generative compression framework that replaces Gaussian codebooks with implicit randomized Hadamard dictionaries. The fast Walsh-Hadamard transform allows for implicitly synthesizing full-dimensional dictionaries with $O(d \log d)$ compute for projection and reconstruction and $O(d)$ memory, without materializing the full-dimensional dictionary. We further introduce an efficient table-free lexicographic coding scheme that makes combinatorial encoding practical at larger dimensions. Beyond images, we adapt the framework to an autoregressive video flow model and develop a streaming compression scheme that periodically re-encodes recently reconstructed frames into a fresh KV-cache prefix, preserving temporal context across segments at no additional transmitted bits. This enables long-form, close to real-time generative video compression on a single GPU. Across image and video benchmarks, H-DDCM achieves competitive rate-distortion-perception performance while substantially reducing the memory and computational overhead of codebook-based training-free generative compression.

Abstract: Task-oriented communication improves communication efficiency by transmitting semantic information required for downstream tasks, but may incur redundancy with previously transmitted information. To address this issue, we propose a variational directed information bottleneck joint source-channel coding (VDIB-JSCC) framework that transmits only previously unacquired information. Specifically, we use directed information to characterize the causal flow of task-relevant incremental information and theoretically demonstrate its advantages over mutual information and conditional mutual information. Based on this metric, we formulate a directed information bottleneck (DIB) objective that balances task-relevant information preservation and redundant information compression. We further derive a variational formulation, termed VDIB, whose tractable upper bound serves as the training loss. The encoder integrates a Vision Transformer for frame-wise semantic extraction and a Mamba-based causal state-space model for capturing temporal dependencies and extracting incremental information, while the decoder uses a Mamba-based state-space model for task inference. Experiments on CIFAR-10 and CIFAR-100 video streams demonstrate that VDIB-JSCC achieves higher classification accuracy with less transmitted information than state-of-the-art semantic communication methods. These results show that leveraging historical knowledge enables task completion with less information, improving communication efficiency.

Abstract: Recovering visually faithful content from damaged video bitstreams is important for reliable multimedia communication and storage. Existing methods exploit local residuals and non-local RGB references of the input video sequence, but both sources can become unreliable after decoding, while processing additional references increases receiver computation and memory. This paper studies multimodal bitstream-corrupted video recovery (MBSCVR) using RGB and depth information, where pre-transmission depth-based geometry aids reconstruction when the received appearance is corrupted. We propose a novel MBSCVR framework that selects and effectively utilizes a single selected non-local RGB reference per recovery window to achieve higher performance. It integrates a reliability-aware multimodal selector that combines valid appearance with depth compatibility before reference encoding. Depth-Conditioned Reference Adaptation (DCRA) introduces selected RGB evidence into feature completion and content generation, while frame-aligned depth conditions completion queries. On the complete BSCV benchmark with oracle masks, MBSCVR uses one selected reference to achieve 32.30 dB on YouTube-VOS and 28.93 dB on DAVIS, surpassing full-reference B2SCVR by 0.27 and 0.33 dB and improving SSIM on both subsets. Matched receiver-forward measurements show 36.8\% lower latency and 40.8\% lower peak memory. These results demonstrate how compact geometric guidance and adapted appearance processing improve the quality--resource trade-off of multimedia recovery.

Abstract: Latent diffusion models enable high-quality video generation, but video VAE encoding and decoding remain computationally expensive. Existing methods use stage-specific lightweight operators, yet either constrain spatial and temporal computation jointly or remove temporal computation altogether, without distinguishing their compression sensitivities in pretrained Conv3D. To address these limitations, we present LightVAE, a compression framework that exploits redundancy in temporal modeling and network depth. LightVAE exactly reparameterizes eligible causal Conv3D layers into an aggregate spatial response and a difference-driven temporal correction. Based on their distinct compression sensitivities, temporal low-rank compression (TLR) keeps the aggregate spatial kernel dense and applies layer-specific low-rank approximation only to the temporal correction. An algebraically equivalent folded implementation converts the arithmetic savings into practical acceleration. LightVAE further uses width-preserving block pruning to reduce depth and probe-guided feature alignment, guided by feedback from a frozen teacher suffix, to recover quality after compression. The framework also extends to video encoders. Across three video VAE backbones and two GPU platforms, LightVAE consistently accelerates decoding. For the Wan VAE family, it achieves up to $2.24\times$ and $1.11\times$ the decoding throughput of state-of-the-art efficient VAE decoders on NVIDIA H100 and RTX~5090, respectively, with higher DAVIS PSNR and lower LPIPS. On MiniMax-H3, it achieves up to $2.97\times$ the original decoding throughput while retaining 99.51\% of the original model's overall VBench-2.0 score. Our code is available at https://github.com/ModelTC/LightVAE.

Abstract: Video diffusion models can reconstruct visually compelling content from severely compressed representations at ultra-low bitrates, but they impose a rigid deployment trade-off. Bidirectional (BI) models achieve high quality through dense full-sequence attention, yet require the entire sequence before reconstruction and incur substantial latency. Autoregressive (AR) models enable low-delay streaming, but their restricted temporal context reduces compression efficiency. This tension is particularly relevant when a video stream must be decoded in real time yet is also stored for later, quality-oriented playback, making AR and BI models natural choices for the live and replay stages, respectively. To support both stages, we present UniGVC, a Unified Generative Video Codec that supports two decoding modes from a single bitstream. UniGVC-BI employs a BI-DiT with dense attention for quality-oriented random-access decoding, whereas UniGVC-AR uses windowed AR-DiT for low-delay streaming. Both paths share a hierarchical latent codec that recurrently produces the same conditions with ultra-low bitrates, allowing UniGVC-BI to serve as a conditional teacher for single-step distillation of UniGVC-AR. A tiny conditional video decoder further reduces streaming latency while preserving most of the reconstruction quality. On standard 1080p benchmarks, UniGVC-BI achieves the best overall rate-distortion-perception trade-off among the evaluated methods and produces temporally stable video at rates as low as 0.0005 bits per pixel. Meanwhile, UniGVC-AR remains competitive while decoding 1080p video at 36.85 FPS, over $250\times$ faster than UniGVC-BI. Code and models will be released later.

Abstract: To stream long videos while maintaining visual quality and temporal consistency, Self Forcing mitigates exposure bias through self-rollout training on self-generated histories with key-value (KV) caching. To keep memory manageable, it detaches historical caches, preserving forward dependencies between chunks but severing the backward gradient paths. We introduce Connected Self Forcing, a training framework that reconnects gradient paths across autoregressive chunks, allowing feedback from later predictions to guide how earlier context is generated. These connections go beyond historical KV-writing: gradients pass through generated latents into the computations that produced them, linking the generation of earlier context to its use in later predictions. To make this connected training memory-efficient, we develop shortcut gradient replay, which recovers cross-chunk gradients without retaining the full rollout computation graph. Integrated with distribution matching distillation, Connected Self Forcing trains historical chunks according to both their direct supervision and their contribution to subsequent generation. Experiments on autoregressive video generation show improvements in long-horizon visual quality and temporal consistency, without changing the inference procedure.

Abstract: Autoregressive video diffusion enables causal video streaming without a bidirectional pass over the full clip, but existing few-step systems usually retain only the opening and most recent frames in a fixed-size KV cache. Once an event leaves this window, later frames can no longer attend to it, a failure we term mid-horizon forgetting. We present Memory Forcing, a few-step streaming method that preserves this missing history without increasing the cache size. Its Archive \& Working Banks partition the cache into sink, archive, and working regions, retaining diverse intermediate events alongside recent motion under fixed memory. Because absolute temporal indices drift outside the training range, Bank-aware RoPE reassigns indices at attention time so each bank remains distinguishable. At 1.3B, Memory Forcing leads on longer clips, shows the smallest drop from 5s to 60s among methods reporting all four lengths, and preserves subjects and scenes through leave-and-return. The same design scales to Wan2.2 5B, producing more physically plausible, realistic, and dynamic videos and, to our knowledge, the first public 5B model on this forcing line.

Abstract: Real-time video super-resolution requires high spatio-temporal fidelity under strict latency constraints, challenging diffusion models due to their iterative sampling cost and limited temporal coordination. We propose a streaming-aware framework that adapts pretrained single-image latent diffusion models for efficient video super-resolution (VSR) by exploiting the sequential structure of video streams. Our Cross-Step Attention mechanism reuses intermediate denoising features across adjacent frames and diffusion steps, enabling temporal information exchange without explicit temporal modeling. We further introduce Trajectory-Coupled Diffusion Scheduling, which aligns adjacent diffusion states and provides cleaner intermediate representations for cross-step conditioning, improving temporal coherence. These components are integrated into a streaming inference pipeline that incrementally propagates latent states across frames, reducing the effective computational complexity from $O(N \cdot S)$ to $O(N + S)$ for $N$ frames and $S$ diffusion steps. Experiments on REDS4 and YouHQ40-Test demonstrate improved perceptual quality and temporal realism while maintaining frame-wise stability. Our method achieves over 40 FPS at $512 \times 512$ resolution after cold start, enabling real-time VSR without explicit temporal modeling.

Abstract: Causal video diffusion models generate video autoregressively, which suits streaming, interactive, and long-video generation. Under standard training, however, they often yield lower generation quality than bidirectional models of the same size. Many existing approaches address this gap by initializing from or distilling a pretrained bidirectional teacher. We instead train a causal model from an image-model initialization, with no bidirectional video model at any stage. Because this path requires neither a large bidirectional teacher nor a complex distillation pipeline, it is simpler and more scalable. On this path, we find that a causal model trained on ground-truth history becomes strongly dependent on it, so that at inference errors in its own generated history propagate forward. We hypothesize that much of this dependence is unnecessary, because the current input already determines much of what the history provides. We propose Conditional Residual Prediction (CRP), a simple recipe for reducing a model's reliance on a condition: the model first predicts the target without the condition, and the condition may only add a residual on top of this prediction. Applied to history, CRP makes the model predict each chunk from the present as far as it can and use the past only for what the present cannot supply. In controlled experiments, CRP nearly closes the 6.14-point gap to a bidirectional model trained under the same setup. Scaling this recipe, we train Optica, a 2B-parameter causal video model that autoregressively generates 5-second 480p videos and reaches 82.78 on VBench with only about 15M training videos.

Abstract: Generative image compression produces visually plausible reconstructions at low bitrates, yet their behavior as the rate approaches zero remains largely unexplored. When pushed below normal operating rates, representative codecs undergo semantic collapse: rather than gracefully losing source-specific detail, they produce malformed or unrecognizable content. Our analysis identifies two factors. As the bitrate decreases, reconstruction losses increasingly conflict with semantic objectives on gradients and visual results, while pixel-space and reconstruction-oriented VAE diffusion models become less efficient on semantic preservation. Guided by these findings, we introduce RAE-CoD, a compression-oriented diffusion (CoD) built in a representation autoencoder (RAE) space with direct alignment between compressed and source representations, preserving recognizable, naturally structured content for a $256\times256$ image with as few as 16 bits. We evaluate this framework using five vision foundation models (VFM) and a blinded vision-language model protocol. On MSCOCO-30K, RAE-CoD stands out from all evaluation. At 0.001-0.008 bpp, it reduces relative VFM feature MSE and Fréchet Distance ratio by at least 25.7% and 69.1% over the best competitors. Meanwhile, semantic recognizability and quality of the reconstructions remain nearly constant while source consistency falls smoothly, replacing abrupt semantic collapse with a graceful transition toward unconditional generation. Code will be released at https://github.com/LuizScarlet/RAE-CoD.

Abstract: Video semantic communication has attracted increasing attention as a promising approach to improving video transmission efficiency. However, most existing approaches rely on computationally intensive deep learning-based video encoders and decoders, which hinders their deployment in resource-constrained scenarios. To address this issue, we propose a lightweight semantic-aware joint source-channel optimization (SAJSCO) scheme that can be integrated into existing digital video communication systems as a plug-in module. Specifically, we develop a video communication system model in which the transmitter jointly optimizes source and channel coding parameters based on the inter-frame semantic importance of the input video and estimated channel state information. On this basis, we formulate an optimization problem that maximizes semantic importance weighted video reconstruction quality under a maximum bitrate constraint. To solve it, we first quantify inter-frame semantic importance using a cosine similarity-based metric with a shifted window mechanism. We then develop a multi-actor proximal policy optimization (MPPO) algorithm to solve the formulated problem by jointly adapting the source compression rate and channel coding rate. The learned policy can be directly applied to different video encoders without encoder-specific retraining or fine-tuning. SAJSCO achieves Bjøntegaard Delta rate reductions of 34.86\% and 18.01\% when integrated with H.265, a conventional video encoder, and DCVC-RT, a deep learning-based video encoder. Over-the-air experiments on a hardware testbed further demonstrate a PSNR gain of up to 1.448 dB with H.265 and an LPIPS reduction of up to 0.033 with DCVC-RT compared with the respective best-performing fixed-parameter baselines.

Abstract: Autoregressive video diffusion enables interactive streaming generation, but suffers from error accumulation over long rollouts. Self-rollout training reduces exposure bias, yet finite rollouts leave long-range drift unresolved. We observe that the noise level of the history key-value (K/V) representations trades visual quality against motion, and that restoring gradients through the history aligns causal training far more closely with bidirectional training. Motivated by these observations, we introduce Self-Aligned Forcing (SAF), a training scheme that aligns the history of each block with the noise level of the block being denoised. Specifically, the history is the K/V produced by preceding blocks at the same denoising stage, so all blocks at a stage can be denoised in a single forward pass under a causal mask. This keeps the noisy history differentiable, allowing future losses to optimize how it is encoded. SAF therefore avoids a separate no-gradient rollout and per-block timestep-zero recaching, training up to 1.8x faster than prior methods with lower memory. At inference, SAF achieves the highest single-GPU throughput among existing methods and keeps one history bank per stage for a multi-GPU pipeline, reaching 49.1 FPS on 4 GPUs. Experiments show superior long-horizon generation with a better balance between visual quality and motion. Project page: https://anonymous.4open.science/w/self-aligned-forcing/.

Abstract: Real-time diffusion-based video super-resolution (VSR) is in high demand for online streaming, yet stringent latency requirements often compromise generative fidelity. We propose ReCaVSR, a Wan2.2-based, one-step framework for streaming VSR that builds on two observations: recycled SR latents retain local temporal context, reducing the need for full historical Key-Value (KV) caches; and individual transformer layers benefit from distinct temporal scopes. ReCaVSR combines three complementary designs: (i) layer-wise cache routing with recycled SR latents: each DiT layer learns its KV-cache temporal scope under a cache budget and exports a static inference schedule, while recycled SR latents propagate local context by conditioning each new block on the model's own preceding predictions. (ii) Multi-Scope Query (MSQ) Discriminator: a compositional discriminator combining global, spatial-window, and temporal-tube feedback for holistic realism, local texture generation, and temporal stability. (iii) LR-conditioned adaptation of FlashDecoder: a VAE decoder that incorporates LR observations for efficient latent decoding. ReCaVSR enables streaming VSR without iterative sampling or full historical KV-cache materialization. Experiments on synthetic and real-world VSR benchmarks show better perceptual quality, temporal consistency, and streaming efficiency than representative VSR baselines. At $1080{\times}1920$ output resolution on a single NVIDIA A100-80GB, ReCaVSR achieves 21.20 FPS with 15.16 GB peak allocated GPU memory, running 2.72$\times$ faster while using 38.0\% less peak allocated memory than FlashVSR Tiny. The code is available at https://github.com/kopperx/ReCaVSR.

Abstract: Under the AI Flow framework, communication networks distribute intelligence across devices, edge servers, and clouds, and computation at the receiver becomes a resource that can substitute for transmitted bits. Generative video compression (GVC) embodies this exchange by sending compact tokens with ultra-low bitrate and letting a generative decoder synthesize the video, yet how much bandwidth savings a unit of decoder compute actually achieves has never been quantified. To fill this vacancy, we model reconstruction quality as a two-factor power law in data rate and decoder compute, which fits measured DISTS of two GVC decoders with a mean error below 3%, and define the information capacity (IC) as the negative logarithmic slope along an iso-quality contour, namely the fraction of rate saved per fractional increase in compute at identical quality. IC is dimensionless and unit-invariant, thus enabling an architecture-agnostic comparison. It forms a field over the operating plane, locating where additional denoising steps are worth their cost. Across five datasets, the 14B decoder trades more compute for fewer rate about ten times more efficiently than the 1.3B decoder. IC also varies significantly across datasets, indicating imbalanced performance on the rate-compute trade-off in GVC methods.

Abstract: Recent video-language models increasingly adopt hybrid architectures that interleave linear and full attention layers for efficient long-context processing. While the recurrent state of linear attention remains fixed in size, the KV cache of full attention continues to grow with the video stream, making eviction necessary under a bounded memory budget. The key challenge in streaming is that eviction must occur before the question arrives, so what to retain has to be decided without the question. Existing eviction methods derive token scores from the KV cache itself, using position, attention, or key-value representations, and attention-based scores further require proxy queries or extra computation. Hybrid backbones offer another source of signal. In gated-delta linear attention, the recurrent state is updated by the residual between each input and what can already be retrieved from the state, so its change over a chunk of frames reflects how much new information the chunk brings. We propose DeltaS, a query-agnostic, training-free method that retains video chunks inducing larger normalized state change, or state drift. In a controlled comparison with the budget and retention policy held fixed, state drift outperforms position-, attention-, and key-value-based signals. With a signal costing only 1.9% of the forward pass, DeltaS surpasses the strongest query-agnostic bounded-memory baseline by 2.1 points on average across six long-video benchmarks and by 5.6 points on the longest benchmark. These results suggest that the two memories of hybrid architectures can work cooperatively. Code is available at https://github.com/MaumAI-Company/DeltaS.

Abstract: The timeliness and accuracy of edge-based video analytics can be hindered by drastic reductions in available bandwidth (ABW) at last-mile edge routers, causing prolonged queuing delays. This work proposes Lizard, a system that leverages video-content-aware packet discarding to mitigate the negative effects of drastic ABW degradation that may frequently occur at a last-mile edge router by judiciously discarding packets that contain frame blocks less important to the analytics at the destination. To achieve this, we first devise a frame-block-aware RTP header extension to effectively decouple packet dependencies to encode frame blocks. Second, Lizard uses a priority-based feedback mechanism that dynamically evaluates packet priorities based on relative accuracy impacts. Third, we develop an adaptive phase-transition-based packet discarding strategy at the router to discard packets that represent unimportant blocks. Our evaluation of Lizard shows improvements over existing methods are substantial: 53.2% reduction in latency and 27.1% increase in analysis accuracy.

Abstract: This paper presents a comprehensive experimental video quality assessment in 5G-based Fixed Wireless Access (FWA) networks, leveraging a real open source 5G network testbed. Addressing the critical need for robust UHD video delivery, the study systematically investigates the impact of transmission modes (SISO and MIMO), modulation schemes (64-QAM and 256-QAM), and network load (one to five users) on both network throughput and perceived video quality. Unlike simulation-based approaches, our methodology integrates automated streaming campaigns with objective full-reference Quality of Experience (QoE) metrics (PSNR, ssim, and VMAF) computed from 4K video streams delivered via MPEG-DASH, enabling frame-level correlation between network performance and user experience. Key findings demonstrate that MIMO consistently outperforms SISO, sustaining higher throughput and superior visual quality even under increasing user demand. Specifically, MIMO with 256-QAM maintained VMAF scores above the visually lossless threshold with five concurrent users, while SISO configurations experienced significant degradation. These results provide empirical evidence for the critical role of spatial diversity and higher-order modulation in ensuring reliable UHD video delivery over FWA. The proposed reproducible framework for detailed quality assessment offers valuable insights for future optimization of 5G FWA deployments, contributing to the practical understanding and deployment strategies for high-quality video services in Open RAN environments.

Abstract: The key-value (KV) cache has become the dominant consumer of memory in large language model (LLM) serving systems as context lengths, concurrency, and request lifetimes grow. High-bandwidth memory (HBM) provides the bandwidth attention decode needs but limited capacity, while off-package memory and storage add capacity but lack the bandwidth to sustain attention decode. High-Bandwidth Flash (HBF) is a promising substrate that combines terabyte-scale capacity with near-HBM read bandwidth. Limited write endurance makes read-only model weights its natural use, but we argue that HBF paired with HBM as a hierarchy can also hold the KV cache. Unlike prior hierarchies, whose secondary tiers are bandwidth bottlenecked, the comparable bandwidths let the two act as one logical memory for the long-context KV cache. HBF capacity enables long-context serving, and sparse attention makes it efficient by limiting KV-cache reads during memory-bound decode. Since HBF reads full flash pages and aggregates bandwidth by accessing thousands of parallel flash planes, sparse attention must be co-designed with these physical properties. We present SPLASH, an algorithm and architecture co-design that virtualizes the KV cache across HBM and HBF and adapts sparse attention to HBF's page granularity and plane-level parallelism. Across models and context lengths, SPLASH improves decode throughput per GPU by 3.5x-11.4x over the evaluated baselines under a 100 ms per-token latency objective, while keeping accuracy within 4% of dense attention across long-context suites.

Abstract: Dynamic ad insertion comparisons can conflate trigger, reach, readiness, playback, billability and measurement even when the accounting is arithmetically correct. We separate these events with an observed-event ledger, a candidate-invariant reference deadline and pod-level contribution accounting. The deadline rule is fixed before candidate assignment and tests whether an admissible transition state remains valid, not whether preparation merely finished earlier. A restricted monotone-playback representation states when media-position summaries suffice; a counterexample shows why they fail over a wider path class. An offline synthetic study exercises the definitions over nine short-lifetime conditions informative for the readiness comparison and nine long-lifetime conditions serving as analytic controls. Across 45,000 shared scripts, two trigger policies share imposed playback paths, latency draws and hypothetical value and cost coefficients. Playhead summaries substantially misclassify reach events in the nonmonotone mixtures, yet neither shortcut reverses the contribution contrast in this grid, because some errors cancel under the shared design. Replacing validity at the deadline with completion by the deadline reverses the comparison in three of the nine informative conditions, all at one of the three latency settings. Scoring readiness at actual viewer arrival rather than at the reference deadline shifts pause-path readiness but changes no contribution sign. These outcomes are consequences of the event definitions applied to established misclassification mechanisms. Conservative bounds retain uncertainty when records are missing, and the artifact records code, seeds, event histories and checking procedures. The evidence is synthetic, uses no commercial telemetry, and ranks neither server-side nor server-guided insertion.

Abstract: Conventional video metrics such as PSNR, SSIM, and VMAF measure visual distortion or perceptual quality, but they do not directly capture semantic preservation: whether compression retains a video's objects, actions, and temporal narrative. Existing Quality-of-Experience (QoE)-driven bitrate-selection and resource-allocation methods primarily aim to minimize rebuffering and bitrate switching while maximizing perceptual video quality, without explicitly considering semantic preservation. To address this gap, we introduce video semantic fidelity (SF), a metric that quantifies how well a compressed video preserves the semantic content of its source. An offline multimodal large language model (MLLM) generates structured descriptions of the reference and compressed versions of the video, and a separate text-only large language model (LLM) evaluates their semantic correspondence. The resulting content-dependent SF--bitrate profiles are cached and queried by the online bitrate selector without invoking MLLMs at runtime. Evaluations on three subjective QoE benchmarks show a consistent positive association between SF and mean opinion scores (MOS). A separate human semantic-rating study evaluates semantic preservation and shows that SF correlates more strongly with human judgments than conventional video metrics. We then embed these profiles into a 5G MEC-assisted video-on-demand (VoD) resource-allocation framework at the base station. When wireless resources cannot support high bitrate levels for all users, the framework uses the SF--bitrate profiles to select bitrate levels jointly across users and reduce the semantic loss caused by the required bitrate reductions. NS-3 simulations with the 5G NR module show that the proposed framework achieves higher average and worst-user SF than the evaluated baselines, with a widening advantage as the wireless resources available to each user decrease.

Abstract: Streaming 3D reconstruction demands both speed and temporal fidelity, goals that existing methods undermine by updating every Gaussian every frame, even in static regions. We present DecoGS, a method for efficient online training of 3D Gaussians from streaming videos. Unlike prior methods that update the entire scene indiscriminately, DecoGS introduces an adaptive mechanism that selectively focuses optimization on spatiotemporal regions exhibiting motion or photometric changes. This targeted training strategy eliminates redundant updates that cause flickering and drift in nominally static regions, while enabling fast, high-fidelity scene updates. The pipeline further integrates region-aware Gaussian management through gradient gating and efficient visibility filtering to maintain temporal coherence and a compact memory footprint. On N3DV and MeetRoom, DecoGS achieves 34.55 and 31.60 dB PSNR respectively, outperforming all streaming and offline baselines, while rendering at 261 FPS with $70\times$ lower temporal flicker than the best prior method, requiring no large-scale pretraining.

Abstract: Leveraging Large Vision-Language Models like CLIP has recently set new benchmarks for No-Reference Image Quality Assessment (NR-IQA). However, the contrastive pretraining of CLIP inherently prioritizes semantic invariance, which often suppresses subtle perceptual signals, a phenomenon we term perceptual submergence. Furthermore, standard preprocessing techniques (e.g., cropping and interpolation) further exacerbate the loss of critical high-frequency quality cues. In this paper, we propose the Cross-modal Perception Alignment Adapter (CMPA), a manifold-aware framework designed to disentangle perceptual distortions from dominant semantics. CMPA introduces a Perception-Sensitive Feature Extractor (PFE) that projects CLIP features into a compact, low-dimensional subspace, explicitly magnifying distortion-induced off-manifold deviations. Subsequently, a Cross-Modal Perception Alignment Injector (PAI) aligns these features with quality-aware text anchors and re-injects them into the backbone. To ensure input fidelity, we also devise a Residual-enhanced Perceptual Downscaling strategy that adaptively compensates for resolution-induced information loss using Just Noticeable Difference (JND) guided frequency re-injection. Extensive evaluations on several benchmark datasets demonstrate that our approach significantly outperforms state-of-the-art methods, effectively recovering the perceptual signals submerged in semantic-dense representations.

Abstract: Benefiting from the powerful generative priors of diffusion models, diffusion-based real-world image super-resolution (Real-ISR) methods have demonstrated impressive performance.To achieve efficient Real-ISR, several recent works have designed one-step diffusion-based models.Howerver, unmediatedly feeding LR into a diffusion model creates a distributional gap with the model's original input.A straightforward approach to reduce the distribution gap is to introduce noise to the LR latents. However, directly adding noise inevitably corrupts the content of the LR images.In this study, we propose DNF-SR, a Dual-input and Negative-aware Feature fine-tuning method for Real-ISR.Specifically, we use a dual-input strategy that concatenates the original LR image with the noisy LR input and feeds them into a diffusion-based image editing model, ensuring both high-fidelity one-step super-resolution and improved perceptual and content consistency.Additionally, the noise present in the noisy LR input introduces randomness and diversity into the outputs. We exploit this property and propose a post-training optimization method, Negative-aware Feature Fine-Tuning (NF2T), which guides the model toward producing higher-quality results.NF^2T classifies multiple outputs into positive and negative subsets and then defines implicit policy improvement directions in both the image and feature spaces, thereby further enhancing the stability of the optimization.Extensive experiments show that DNF-SR outperforms other methods.Code will be released.

Abstract: Traditional recommender systems are typically trained to predict what item users will interact with next, but not why. However, offering personalized evidence for why a user might like the predicted item is an important way to enhance the service and to raise the likelihood that the user will be genuinely interested in the recommendation. This service can be delivered by integrating a frontier-model call into the member-facing pipeline, but it will add extra cost and latency. In this paper, we train a recommender LLM to generate personalized explanations for its reccomendation, based on the user's watching history at a large video streaming service. We impose two requirements on the generated explanation: it must be faithful to the elements of the shows it links, and it must be strictly non-harmful to the user. To this end, we first train two LLM-judge reward models covering three specific criteria, and propose constrained GRPO to incorporate these different criteria. On a held-out real-world testing set, our fine-tuned model improves the all-three-criteria PASS rate rises from 0.649 to 0.956 under our own judges and from 0.677 to 0.931 under an independent judge, where as the frontier generator performs similar to the untuned recommender baseline. We conduct further experiments to show that the model's language and recommendation abilities remain unchanged. Based on these results, we conclude that an LLM-based recommender can be fine-tuned on other complex tasks without compromising its original recommendation performance, thus provide insights for further agentic user interface powered by a single model.

Abstract: In this paper, we study sequential lossy compression under a causal conditional perception criterion comparing source and reconstruction distributions given the same reconstruction history. For first-order Markov sources, we formulate the finite-horizon nonanticipative rate-distortion-perception function (NRDPF) with stagewise constraints and establish one-shot lower and upper bounds on the minimum variable-length sum rate using a strengthened strong functional-representation lemma (SFRL) and common randomness. For time-varying scalar Gauss--Markov sources under pointwise mean-squared error (MSE) and conditional squared Wasserstein-$2$ fidelity, we prove Gaussian optimality, derive a log-variance characterization, and obtain a closed-form solution that recovers the classical Gaussian nonanticipative rate-distortion function (NRDF) when perception is unconstrained and the classical Gaussian RDPF when the source is stationary and memoryless.