AI research tackles long-video memory with novel weight-internalization techniques · 7 sources tracked
ByPulseAugur Editorial·[109 sources]·
Researchers are developing novel methods to enhance the memory capabilities of AI models, particularly for processing long video streams. Several papers introduce techniques that internalize historical data into model weights or abstract states, rather than relying solely on context windows. These approaches aim to improve consistency, reduce computational overhead, and enable more sophisticated video generation and understanding across various benchmarks.
AI
IMPACT
These advancements could lead to more coherent and efficient AI video generation and understanding systems, impacting fields from content creation to robotics.
RANK_REASON
Multiple research papers introducing novel methods for AI memory in video generation.
arXiv:2610.23542v1 Announce Type: cross Abstract: A memory test for a video world model should say whether the scene was set, whether the asked change was carried out, and whether the last frame shows the result. Current tests fold these into one number, so a low score is easy to…
arXiv:2610.22070v1 Announce Type: cross Abstract: Streaming vision-language models (VLMs) must continuously integrate information from the past while remaining responsive to current observations and user queries. Existing memory mechanisms typically compress, retrieve, or cache h…
arXiv:2610.20822v1 Announce Type: new Abstract: Text-to-video models often generate the visual concepts requested in natural language prompts but fail to satisfy their intended ordering, timing, and persistence requirements. We propose representing these requirements as specifica…
PPMem is a video world model that uses compact plane primitives as persistent 3D memory to generate long coherent videos with explicit metric geometry.
MIST separates scene setup from change execution to isolate memory failures in video world models, and M3-Agent improves instruction execution to significantly boost benchmark scores.
arXiv:2605.08158v2 Announce Type: replace-cross Abstract: Long video understanding with multimodal language models suffers from three compounding bottlenecks: heavy decode cost to obtain dense RGB frames, token growth with frame count, and weak motion perception under sparse keyf…
arXiv cs.AI
TIER_1English(EN)·Hang Xu, Guohui Zhang, Xiaoxiao Ma, Hu Yu, Jie Huang, Jiayin Xiao, Lin Song, Haoyang Huang, Nan Duan, Feng Zhao·
arXiv:2610.16632v1 Announce Type: cross Abstract: Memory-based long-video generation aims to combine user control with long-range visual consistency. We organize this process into three functional components: a planner, an adapter, and a generator. Representative multi-shot syste…
arXiv:2610.17434v1 Announce Type: cross Abstract: Recent advances in video world models have enabled plausible interactive environments, yet long-horizon generation places a fundamental demand on visual memory: the previously observed world must remain consistent as the model con…
arXiv:2610.17926v1 Announce Type: cross Abstract: Autoregressive video generation relies on memory retrieval to model long-range dependencies. This capability becomes particularly critical for multi-shot generation, where models must reconcile substantial visual changes across sh…
arXiv cs.LG
TIER_1English(EN)·Lorenzo Venturoli, Nedko Savov, Xi Wang, Marc Pollefeys, Luc Van Gool·
arXiv:2610.19795v1 Announce Type: cross Abstract: Video world models aim to simulate complex environments with realistic appearance and dynamics. Maintaining consistency over long interactions requires tracking past actions and understanding how those actions affect the environme…
MaLoW internalizes streaming video history into evolving LoRA weights via a recurrent HyperNetwork, improving streaming vision-language performance with minimal latency overhead.
Speculative Forcing accelerates few-step autoregressive video generation by accepting early token drafts and skipping unnecessary denoising steps while preserving output quality.
Representation Forcing improves few-step video generation by using feature-space supervision instead of trajectory regression, boosting quality metrics over causal ODE baselines.
Speculative Sampling enables interactive video generation with large bidirectional windows by caching predictions and delivering them incrementally, reducing latency without sacrificing quality.
Hour-long videos, streaming feeds, and movie-length content are pushing vision-language models (VLMs) past their breaking point -- the standard recipe of encoding each frame into visual tokens yields sequences that grow linearly with video length, causing prohibitive memory consu…
arXiv:2610.12104v1 Announce Type: cross Abstract: Building a capable video editor remains significantly harder than a video generator: editing requires (source, instruction, edited) triplets that are prohibitively expensive to annotate and difficult to synthesize at scale, wherea…
arXiv:2610.09328v1 Announce Type: cross Abstract: Multi-subject image generation requires rewards that verify whether requested attributes, actions, and relations hold for the specified reference subjects. Subject presence alone does not establish that the correct subjects partic…
Generating an egocentric video from a single exocentric recording is a challenging case of novel view synthesis, as the two cameras share little overlap and much of the target view is unobserved. Current state-of-the-art methods reconstruct the scene explicitly by estimating dept…
Video-language models commonly assume that more temporal observations lead to more reliable reasoning. We question this assumption and argue that the key challenge is not merely processing more video frames efficiently, but learning to reason reliably under limited temporal evide…
arXiv:2407.07111v2 Announce Type: replace-cross Abstract: The rapid development of diffusion models (DMs) has significantly advanced image and video applications, making "what you want is what you see" a reality. Among these, video editing has gained substantial attention and see…
We introduce QuadTok, a novel framework for visual tokenization and autoregressive image generation. Compared to traditional approaches using 2D grids or 1D token sequences, we propose a hierarchical quadtree structure, bridging the gap between 2D spatial binding and 1D sequence-…
Video-language models commonly assume that more temporal observations lead to more reliable reasoning. We question this assumption and argue that the key challenge is not merely processing more video frames efficiently, but learning to reason reliably under limited temporal evide…
Autoregressive video generation requires denoising the current frames while writing their key-value representations as context for future predictions. However, these two roles typically share parameters, and we find that their gradients exhibit distinct patterns and systematic ne…
Highly compressed video autoencoders offer an effective way to accelerate video diffusion models, as the Diffusion Transformer (DiT) operates on far fewer tokens. However, such autoencoders are challenging to train, since a higher compression ratio degrades reconstruction quality…
Interactive video world models need to generate each video chunk efficiently while responding faithfully to user controls. Many systems use chunk-wise autoregressive generation with few-step denoising, but each chunk still requires several costly denoising iterations. Training-fr…
Interactive video world models need to generate each video chunk efficiently while responding faithfully to user controls. Many systems use chunk-wise autoregressive generation with few-step denoising, but each chunk still requires several costly denoising iterations. Training-fr…
Bidirectional video diffusion models denoise entire videos in parallel, yet when trained on effectively unlimited in-distribution data from procedural generators, continue to violate physical laws and simple symbolic rules. We introduce Serial-to-Parallel Diffusion (S2PD), which …
Rapid progress in video generation has led to a plethora of models that differ substantially in capability and generation cost. This raises a natural question: can each request be efficiently routed to an appropriate model? We find that even when the consensus of the other annota…
arXiv cs.AI
TIER_1English(EN)·Huawei Lin, Tony Geng, Zhaozhuo Xu, Weijie Zhao·
arXiv:2505.13439v2 Announce Type: replace-cross Abstract: Autoregressive (AR) models have recently shown strong performance in image generation, where a critical component is the visual tokenizer (VT) that maps continuous pixel inputs to discrete token sequences. The quality of t…
arXiv:2311.18837v2 Announce Type: replace-cross Abstract: Diffusion models have achieved significant success in image and video generation. This motivates a growing interest in video editing tasks, where videos are edited according to provided text descriptions. However, most exi…
arXiv cs.AI
TIER_1English(EN)·Ziqi Ma, Shreya Sharma, Mohamed El Banani, Katja Schwarz, Chongjie Ye, Chao-Yuan Wu, Li Fei-Fei, Ben Mildenhall, Georgia Gkioxari, Justin Johnson, Gowthami Somepalli·
arXiv:2610.03636v1 Announce Type: cross Abstract: Camera-controlled video models are rapidly advancing toward long generation horizons and complex camera control. A key failure mode is 3D inconsistency: as the camera moves, objects lose permanence and scene structures shift. Exis…
arXiv:2610.03510v1 Announce Type: cross Abstract: Recent advances in autoregressive video generation have improved temporal consistency over extended durations, yet interactive storytelling requires more than continuous scene extension: a new shot may combine characters and backg…
Autoregressive video diffusion supports streaming generation and interactive control, but its KV cache grows continuously with the generated history. Existing compression strategies either discard history using fixed windows or select tokens through local attention and similarity…
Bidirectional video diffusion models denoise entire videos in parallel, yet when trained on effectively unlimited in-distribution data from procedural generators, continue to violate physical laws and simple symbolic rules. We introduce Serial-to-Parallel Diffusion (S2PD), which …
We present Kandinsky 6.0 Video, a family of foundation diffusion models for synchronized text-to-audio-video generation, comprising Kandinsky 6.0 Video Lite (3B parameters) and Kandinsky 6.0 Video Pro (29B parameters). Both models generate 5-second video clips with synchronized 4…
Efficient long-video understanding with vision-language models (VLMs) is often framed as selecting informative frames or visual tokens at a fixed native resolution. We show that per-frame resolution can instead be traded for denser temporal coverage, while front-end decoding late…
arXiv cs.LG
TIER_1English(EN)·Chi Zhang, Shi Haoyang, Yueyi Liu, Ruichuan An, Junkang Zhou, Chang Li, Xiuyuan Lu, Yichi Zhang, Bo Wang, Yuhang Wu, Sen Cui, Miao Liu·
arXiv:2609.35763v3 Announce Type: replace Abstract: Distributional training provides collective supervision for one-step visual generation by matching real and generated features in frozen representation spaces. We introduce a unified theoretical framework that separates distribu…
Modern autoregressive (AR) video diffusion models excel at short-horizon video generation, yet generating long videos remains challenging due to drifting, where colors and textures shift, and motion dynamics decay. Existing works primarily rely on KV conditioning, which selects o…
Streaming video generation has benefited from distribution matching distillation (DMD), which matches the joint distribution of video frames to a video teacher's approximation of the real video distribution. Although this joint matching mitigates drift during autoregressive rollo…
arXiv cs.AI
TIER_1English(EN)·Amine Ouasfi, Runjia Li, Junlin Han, Eric Marchand, Philip H. S. Torr, Adnane Boukhayma·
arXiv:2609.39504v1 Announce Type: cross Abstract: We present PartiCam, a training-free Particle filtering rooted method for improved Camera controlled video generation. Generating videos that follow a precisely specified camera trajectory remains challenging for large video diffu…
arXiv:2609.37925v1 Announce Type: cross Abstract: Autoregressive (AR) video diffusion enables low-latency, streamable video generation, but prediction errors often accumulate over long rollouts. Training the generator on its own rollouts exposes it to these imperfect histories. H…
arXiv cs.AI
TIER_1English(EN)·Chi Zhang, Yueyi Liu, Haoyang Shi, Ruichuan An, Haoyu Li, Yuhang Wu, Sen Cui, Miao Liu·
arXiv:2609.35491v2 Announce Type: replace-cross Abstract: Few-step autoregressive video generation commonly relies on Distribution Matching Distillation (DMD), requiring a bidirectional diffusion teacher and an online fake-score model. We instead learn the rollout distribution di…
Video generation has rapidly progressed from short, low-quality clips to high-resolution, long-duration sequences with complex spatiotemporal dynamics. Despite strong generative priors learned through large-scale pretraining, pretrained video models often fail to reliably follow …
Long-horizon video generation requires models to effectively leverage an increasingly long generation history. As the generated history grows, retaining all previous content becomes increasingly expensive and redundant, making effective historical selection essential. Existing ap…
Recent video-based world models pair the scalability of autoregressive (AR) prediction with the visual quality of diffusion models. The choice of scene tokenizer is paramount for the optimal performance of each of these, both in terms of fidelity and semantics. Flexible-length, c…
Video diffusion models are increasingly developed into specialized models for diverse downstream tasks, and this development often includes a distillation stage, for example to accelerate sampling or to improve long-video generation. This stage is typically repeated for every spe…
Autoregressive (AR) video diffusion enables low-latency, streamable video generation, but prediction errors often accumulate over long rollouts. Training the generator on its own rollouts exposes it to these imperfect histories. However, existing video-level distribution matching…
World models, game simulators, and long-take video creation require coherent scene evolution and sustained dynamics over extended durations. Autoregressive (AR) video diffusion provides a natural framework for long-horizon generation, yet extended rollouts often become near-stati…
arXiv cs.CV
TIER_1English(EN)·Zhixiao Wang, Yang Chen, Yifan Xu, Chenbin Li, Yue Chang, Kai Long, Fengming Dong, Limin Wang·
arXiv:2610.22608v1 Announce Type: new Abstract: Recent advances in video generation let users create high-quality videos from natural language descriptions, yet obtaining results that match user intent remains time-consuming: individual generations are computationally expensive, …
arXiv cs.CV
TIER_1English(EN)·Yuze He, Bin Tan, Zelin Gao, Yujun Shen, Yong-Jin Liu, Nan Xue, Yinghao Xu·
arXiv:2610.23795v1 Announce Type: new Abstract: Video world models that generate and explore environments at scale need a persistent 3D memory of what they have produced. Plane primitives offer a natural candidate: a compact set of oriented planes captures the dominant structural…
arXiv:2610.23856v1 Announce Type: new Abstract: Generated videos often lose geometric consistency under camera motion: structures warp under viewpoint change, parallax becomes inconsistent, and smooth camera movement may collapse into shot-like transitions. We argue that this fai…
arXiv:2610.23021v1 Announce Type: new Abstract: In real-world videos, sound events are often associated with localized and fine-grained visual cues, such as object motion, object-part dynamics, contact events, and interactions between objects. Existing video-to-audio (V2A) method…
arXiv cs.CV
TIER_1English(EN)·Nedko Savov, Stefan Maria Ailuro, Danda Pani Paudel, Yilun Du, Xi Wang, Luc Van Gool·
arXiv:2610.23423v1 Announce Type: new Abstract: Video world models must capture how a scene evolves beyond the current camera view to generate consistent observations when its content reappears. Existing approaches rely on observation history, static point clouds, or image caches…
arXiv cs.CV
TIER_1English(EN)·Hyunwook Choi, Seungryong Kim, Dongyoon Han, Sangdoo Yun·
arXiv:2610.22347v1 Announce Type: new Abstract: Few-step autoregressive video generators enable streaming generation, yet they run every token through the final denoising step, regardless of how much refinement it needs. We find this uniform computation is often unnecessary: many…
arXiv:2610.21129v1 Announce Type: new Abstract: Diffusion Transformers (DiTs) dominate video generation but suffer from high inference cost due to the quadratic complexity of self-attention. Sparse attention mitigates this by skipping unnecessary computation in the attention matr…
arXiv:2605.21042v2 Announce Type: replace Abstract: Diffusion models have achieved impressive performance in video generation, but their iterative denoising process remains computationally expensive due to the large number of tokens processed at each timestep. Recently, progressi…
arXiv cs.CV
TIER_1English(EN)·Samuel Teodoro, Agus Gunawan, Soo Ye Kim, Jihyong Oh, Munchurl Kim·
arXiv:2412.12877v3 Announce Type: replace Abstract: Recent Artificial Intelligence (AI)-based video editing has enabled users to edit videos through simple text prompts, significantly simplifying the editing process. However, recent zero-shot video editing techniques focus on glo…
arXiv:2610.19063v1 Announce Type: new Abstract: As video models enable longer and more realistic interactive experiences, maintaining a world consistent with its history becomes essential. This requires both preserving previously observed content and ensuring coherent world evolu…
arXiv cs.CV
TIER_1English(EN)·Eren Cetin, Julius Erbach, Vukasin Bozic, Hyojun Go, Zixiang Zhao, Nikolai Kalischek, Federico Tombari, Konrad Schindler, Dominik Narnhofer·
arXiv:2610.19257v1 Announce Type: new Abstract: Memory for long-horizon video generation must satisfy two competing requirements: it must remain bounded so generation can continue indefinitely, while still recovering what was observed at a place when the camera returns, regardles…
arXiv:2610.20111v1 Announce Type: new Abstract: Distilling pretrained video diffusion models into few-step autoregressive (AR) generators typically combines student initialization with distribution matching distillation (DMD). An intermediate AR teacher can provide causally compa…
arXiv:2610.20556v1 Announce Type: new 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 l…
arXiv:2610.11023v1 Announce Type: new Abstract: Identity-preserving video generation aims to maintain a subject's identity while synthesizing realistic videos. Yet a single reference portrait captures the subject's appearance under only one facial configuration. As expressions ch…
arXiv:2610.11037v1 Announce Type: new Abstract: Recent image editing systems have achieved impressive semantic understanding, visual fidelity, and instruction-following ability, while video editing remains substantially more difficult and costly. In this paper, we present a simpl…
arXiv cs.CV
TIER_1English(EN)·Chengfeng Han, Baole Ai, Xianlu Bian, Jie Yao, Zilong Huang, Ang Wang, Dandan Ding·
arXiv:2610.11302v1 Announce Type: new Abstract: Training-free sparse attention offers a practical acceleration solution to Diffusion Transformers (DiTs) via reducing computations without fine-tuning. It typically involves estimating the importance of query-key regions and derivin…
arXiv cs.CV
TIER_1English(EN)·Wenfang Sun, Yingjun Du, Cees G. M. Snoek·
arXiv:2610.10893v1 Announce Type: new Abstract: Video-language models commonly assume that more temporal observations lead to more reliable reasoning. We question this assumption and argue that the key challenge is not merely processing more video frames efficiently, but learning…
arXiv cs.CV
TIER_1English(EN)·Lan Feng, Peter Karkus, Maximilian Igl, Julius Berner, Yuxiao Chen, Shuhan Tan, Alexandre Alahi, Boris Ivanovic, Marco Pavone·
arXiv:2610.11498v1 Announce Type: new Abstract: Video diffusion and flow models require many sequential evaluations, making generation computationally expensive. Few-step distillation reduces this cost but poses a capacity allocation problem: a student must match the teacher's it…
arXiv:2610.11736v1 Announce Type: new Abstract: Modern video generators can realize increasingly complex visual narratives, positioning the prompt enhancer (PE) as a critical bridge from concise user instructions and multimodal references to structured cinematic plans. However, e…
arXiv:2610.11746v1 Announce Type: new 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-aw…
arXiv:2610.11756v1 Announce Type: new 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 ev…
arXiv:2610.11770v1 Announce Type: new Abstract: AI-Native video creation is shifting from isolated video clips toward iterative video creation workflows. However, existing datasets remain largely video clips, representing video generation and editing as separate tasks rather than…
arXiv:2610.12156v1 Announce Type: new 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, …
arXiv:2610.12382v1 Announce Type: new Abstract: Faithful visual world simulation requires generated videos to maintain 4D world consistency, encompassing both static and dynamic consistency. Static consistency requires coherent 3D structure in static environments across viewpoint…
arXiv cs.CV
TIER_1English(EN)·Suhwan Cho, Yonwoo Choi, Soongjin Kim, Jicheol Park, Taegyu Lim·
arXiv:2610.12442v1 Announce Type: new Abstract: Generating an egocentric video from a single exocentric recording is a challenging case of novel view synthesis, as the two cameras share little overlap and much of the target view is unobserved. Current state-of-the-art methods rec…
arXiv:2609.34221v2 Announce Type: replace Abstract: Despite rapid progress, world models still lack explicit, persistent structural memory, making it difficult to preserve consistent world structure during continual scene expansion and cross-view revisits. To address this limitat…
arXiv cs.CV
TIER_1English(EN)·Zhenghong Zhou, Xiaohang Zhan, Zhiqin Chen, Soo Ye Kim, Nanxuan Zhao, Haitian Zheng, Qing Liu, He Zhang, Zhe Lin, Yuqian Zhou, Jiebo Luo·
arXiv:2603.15614v2 Announce Type: replace Abstract: Recent video diffusion models achieve strong visual quality and temporal coherence, but still lack coordinated control over scene layout, customized subject identity, and camera/subject motion. We study controllable video genera…
arXiv cs.CV
TIER_1English(EN)·Jiyoung Kim, Paul Hyunbin Cho, Jisu Nam, Donghoon Lee, Hyunsung Go, Yeonkyeong Lee, Hansaem Kim, Seungryong Kim·
arXiv:2610.10524v1 Announce Type: new Abstract: Highly compressed video autoencoders offer an effective way to accelerate video diffusion models, as the Diffusion Transformer (DiT) operates on far fewer tokens. However, such autoencoders are challenging to train, since a higher c…
arXiv:2610.10497v1 Announce Type: new Abstract: We introduce QuadTok, a novel framework for visual tokenization and autoregressive image generation. Compared to traditional approaches using 2D grids or 1D token sequences, we propose a hierarchical quadtree structure, bridging the…
arXiv cs.CV
TIER_1English(EN)·Yuxiang Xiong, Ruiyan Wang, Wenqiang Wang, Teng Hu, Songhang Shen, Bohao Feng, Hongqian Deng, Ran Yi·
arXiv:2610.10457v1 Announce Type: new Abstract: Diffusion Transformers (DiTs) achieve remarkable performance in video synthesis, but their iterative denoising process suffers from high inference latency. To address this, caching has emerged as an effective acceleration strategy b…
arXiv cs.CV
TIER_1English(EN)·Zihan Su, Junhao Zhuang, Yaowei Li, Siwen Lu, Haoran Li, Lingen Li, Haoyu Wu, Weiyang Jin, Songchun Zhang, Haoyang Huang, Chun Yuan, Zeyue Xue, Nan Duan·
arXiv:2610.10429v1 Announce Type: new Abstract: Autoregressive video generation requires denoising the current frames while writing their key-value representations as context for future predictions. However, these two roles typically share parameters, and we find that their gradi…
arXiv cs.CV
TIER_1English(EN)·Naomi Ken Korem, Mohamed Oumoumad, Omer Hagage, Amir Gam, Matan Ben Yosef, Harel Cain, Urska Jelercic, Ofir Bibi, Yaron Inger, Or Patashnik, Daniel Cohen-Or·
arXiv:2604.11788v2 Announce Type: replace Abstract: High dynamic range (HDR) imagery offers a rich and faithful representation of scene radiance, but remains challenging for generative models due to its mismatch with the bounded, perceptually compressed data on which these models…
arXiv cs.CV
TIER_1English(EN)·Shangye Song, Dong Gong, Hong Jia, Yun Sing Koh, Xinyu Zhang·
arXiv:2610.08777v1 Announce Type: new Abstract: Interactive video world models need to generate each video chunk efficiently while responding faithfully to user controls. Many systems use chunk-wise autoregressive generation with few-step denoising, but each chunk still requires …
arXiv:2610.08772v1 Announce Type: new Abstract: Diffusion Transformers (DiTs) have achieved strong performance in image and video generation, but the quadratic complexity of full attention makes high-resolution generation computationally expensive. Window attention offers an effi…
arXiv:2610.08684v1 Announce Type: new Abstract: Modern video models can generate realistic videos from real appearance references and proxy renders that specify scene structure, viewpoint changes, and motion. Evaluating this render-to-real capability requires a real target video …
arXiv cs.CV
TIER_1English(EN)·Jeongwoo Shin, Youngyoon Choi, Sangwoo Jo, Hyunmog Kim, Sungjoon Choi, Joonseok Lee, Jaewoong Choi, Jaemoo Choi·
arXiv:2610.03120v1 Announce Type: new Abstract: Modern autoregressive (AR) video diffusion models excel at short-horizon video generation, yet generating long videos remains challenging due to drifting, where colors and textures shift, and motion dynamics decay. Existing works pr…
arXiv:2610.02779v1 Announce Type: new Abstract: In this paper, we present trajectory-aware reuse and adaptive correction (TRAC), a training-free framework for efficient autoregressive (AR) video generation. Existing acceleration methods mainly target single-trajectory generation …
arXiv cs.CV
TIER_1English(EN)·Yunseung Ok (Kyung Hee University), Hyunsoo Kim (The University of Texas at Austin), Minseo Kim (Kyung Hee University), Suhyun Kim (Kyung Hee University)·
arXiv:2610.02914v1 Announce Type: new Abstract: Autoregressive video models can generate minute-long videos in real time, but they produce generic subjects from text rather than specific subjects from user-provided images. Existing customization methods either require costly per-…
arXiv:2610.03543v1 Announce Type: new Abstract: Streaming video generation has benefited from distribution matching distillation (DMD), which matches the joint distribution of video frames to a video teacher's approximation of the real video distribution. Although this joint matc…
arXiv:2610.03221v1 Announce Type: new Abstract: Video creation spans text-to-video (T2V), image-to-video (I2V), and condition-based generation, yet video diffusion models remain costly because they repeatedly evaluate large backbones during sampling. Distribution matching distill…
arXiv cs.CV
TIER_1English(EN)·Zhuo Ning, AmirHossein Naghi Razlighi, Sagi Polaczek, Daniel Cohen-Or, Ali Mahdavi-Amiri·
arXiv:2610.00691v1 Announce Type: new Abstract: Recent joint audio-video generative models can synthesize realistic videos with synchronized sound, but typically generate audio as a single mixed track. This limits source-level control and differs from practical audiovisual workfl…
arXiv cs.CV
TIER_1English(EN)·Yuxin Cao, Wei Song, Shangzhi Xu, Jingling Xue, Jin Song Dong·
arXiv:2602.10639v2 Announce Type: replace Abstract: Video Large Language Models (VideoLLMs) have achieved strong performance on video understanding tasks, yet existing benchmarks evaluate only what models predict, leaving the stability of how they generate largely unexamined. We …
arXiv:2610.02197v1 Announce Type: new Abstract: Video generation models have achieved remarkable visual fidelity and have strong potential to become general-purpose world simulators. Despite this progress, they still fail to generate videos which adhere to laws of physics. The pr…
arXiv cs.CV
TIER_1English(EN)·Yiwen Zhang, Haocheng Xi, Michael Tian-Yue Liu, Alexei A. Efros, Hadar Averbuch-Elor, Qianqian Wang, Haiwen Feng·
arXiv:2610.02153v1 Announce Type: new Abstract: Long-horizon autoregressive video generation is limited by a finite context window. When an object or scene falls out of context, its fine-grained visual details may be lost and difficult to recover upon reappearance. To retain acce…
arXiv cs.CV
TIER_1English(EN)·Cusuh Ham, Fabian Caba Heilbron, Josef Sivic, Bryan Russell·
arXiv:2610.01884v1 Announce Type: new Abstract: We introduce an approach for collection-grounded B-roll sequence generation. Given a user's video collection, a directive given in natural language, and a target duration, the goal is to produce a multi-shot sequence that complement…
arXiv:2610.01661v1 Announce Type: new Abstract: Despite remarkable progress, video generation models often produce highly similar outputs when repeatedly sampled from the same prompt, limiting their usefulness for creative exploration. Existing diversity evaluations primarily rel…
arXiv cs.CV
TIER_1English(EN)·Tongcheng Zhang, Jun Zhu, Jianfei Chen·
arXiv:2610.01052v1 Announce Type: new Abstract: We argue that as video generation extends to longer durations, subject consistency should be evaluated over \textit{dynamic subject sets}. We therefore introduce \textbf{DynSC-Eval}, an evaluation framework that dynamically tracks e…
arXiv cs.CV
TIER_1English(EN)·Jiho Jang, Jinyoung Kim, Nojun Kwak, Kyungjune Kim·
arXiv:2610.01039v1 Announce Type: new Abstract: While recent video generative models can synthesize high-fidelity videos, they struggle to portray plausible physical interactions and the resulting state transitions, a critical bottleneck for applications in robotics and VR/AR. To…
arXiv cs.CV
TIER_1English(EN)·Mikhail Dereviannykh, Vikram Voleti, Simon Donne, Mallikarjun Byrasandra Ramalinga Reddy, Shimon Vainer, Mark Boss·
arXiv:2610.00686v1 Announce Type: new Abstract: Recent video-based world models pair the scalability of autoregressive (AR) prediction with the visual quality of diffusion models. The choice of scene tokenizer is paramount for the optimal performance of each of these, both in ter…
arXiv:2609.38691v1 Announce Type: new Abstract: Streaming video generators allow users to dynamically modulate video synthesis via mid-stream prompt switching. Existing streaming methods can respond to the updated instruction while still cutting corners, prematurely realizing goa…
arXiv:2609.38839v1 Announce Type: new Abstract: Long-horizon video generation requires models to effectively leverage an increasingly long generation history. As the generated history grows, retaining all previous content becomes increasingly expensive and redundant, making effec…
arXiv cs.CV
TIER_1English(EN)·Byoungwoo Park, Jaemoo Choi, Juho Lee, Yongxin Chen·
arXiv:2609.38562v1 Announce Type: new Abstract: World models, game simulators, and long-take video creation require coherent scene evolution and sustained dynamics over extended durations. Autoregressive (AR) video diffusion provides a natural framework for long-horizon generatio…
arXiv cs.CV
TIER_1English(EN)·Lingyu Liu, Yaxiong Wang, Li Zhu, Zhedong Zheng·
arXiv:2609.39132v1 Announce Type: new Abstract: We study few-step video generation, i.e., distilling a multi-step video generator, which typically requires tens of sampling steps, incurring substantial latency and compute, into a few-step student. Consistency distillation is a co…
arXiv:2609.40037v1 Announce Type: new Abstract: Few-step autoregressive video generation enables efficient streaming synthesis, but errors introduced in early temporal blocks are reused as context and can propagate through subsequent rollouts, leading to detail degradation, struc…
arXiv:2609.38154v1 Announce Type: new Abstract: Video diffusion models are increasingly developed into specialized models for diverse downstream tasks, and this development often includes a distillation stage, for example to accelerate sampling or to improve long-video generation…
<!-- SC_OFF --><div class="md"><p>Ive been working on this for years, and FINALLY - THANKS TO OPUS 5.5, this dream is reality <a href="https://youtu.be/FV9YaDkvmTI?si=KkymN0a_i0QIngBh">https://youtu.be/FV9YaDkvmTI?si=KkymN0a_i0QIngBh</a> The presenter in this video isn't me. It's…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1x0xchw/duomatching_jointmarginal_distribution_matching/"> <img alt="DuoMatching: Joint–Marginal Distribution Matching for Few-Step Video Generation" src="https://external-preview.redd.it/d3dkbWZmYng0YXVo…
<table> <tr><td> <a href="https://www.reddit.com/r/StableDiffusion/comments/1wxlyq2/pdmd_projected_distribution_matching_distillation/"> <img alt="PDMD: Projected Distribution Matching Distillation for Video Diffusion Models" src="https://external-preview.redd.it/OWZubGE1ZmFvaHRo…