Pith. sign in

REVIEW 2 major objections 4 minor 34 references

AlayaWorld claims that stable, controllable long-horizon interactive video can be generated at interactive latency by combining a persistent scene anchor, compressed temporal memory, geometry-aligned spatial memory, and drift-aware training

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 15:43 UTC pith:BYKGV5JE

load-bearing objection A competent engineering report with a coherent integrated system; the headline benchmark claim is not yet established because the prompt-adaptation protocol is one-sided as reported, and error bars are absent. the 2 major comments →

arxiv 2607.18367 v1 pith:BYKGV5JE submitted 2026-07-20 cs.AI

AlayaWorld: Interactive Long-Horizon World Modeling -- Full Technical Report

AlayaWorld Team: Kaipeng Zhang , Chuanhao Li , Yifan Zhan , Yongtao Ge , Yuanyang Yin , Jiaming Tan , Kang He , Liaoyuan Fan
show 9 more authors
Mingliang Zhai Ruicong Liu Xiaojie Xu Xuangeng Chu Zhen Li Zhengyuan Lin Zhixiang Wang Zian Meng Zihui Gao
This is my paper
classification cs.AI
keywords interactive world modelsvideo generationautoregressive diffusionlong-horizon stabilitycamera controlspatial memorydiffusion distillationdrift-aware training
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper tries to establish that a video world model can replace traditional game production for explorable, persistently consistent virtual worlds. Its central claim is that four coupled properties—responsiveness to camera and text control, scene consistency on revisits, stability over long roll-outs, and interactive speed—can be achieved together in one autoregressive diffusion transformer. AlayaWorld generates 24-fps video chunk by chunk in latent space, with each chunk produced in four sampling steps rather than about thirty, while keeping a bounded context that does not grow with the horizon. The paper reports leading scores on a public long-horizon interactive-world benchmark on brightness consistency, color-temperature stability, sharpness retention, trajectory accuracy, and memory symmetry. If true, the result matters because it makes personalized, rapidly evolving interactive environments practical without hand-built assets, physics, or gameplay programming.

Core claim

On the paper's own terms, the discovery is that a causal autoregressive factorization of video generation, conditioned on a bounded prefix of four clean context streams, can maintain long-horizon stability and control. At each chunk, the model prepends a fixed global anchor frame (the sink), a compressed embedding of the last six latent frames, a geometry-aligned reprojection of previously seen views into the current camera, and the single most recent frame; full self-attention then denoises only the target chunk. Revisiting a location is supported by an explicit cache of frames with monocular depth and camera poses, rendered into the current view by depth-based reprojection. Drift is contai

What carries the argument

The load-bearing object is the per-chunk prefix token sequence S_i = (sink; temporal memory; spatial memory; nearby; target). The sink is a single clean latent frame pinned at temporal position 0 and held fixed across all chunks as the global identity anchor; temporal memory is a compressed embedding of the last six latent frames; spatial memory is a depth-based reprojection of cached frames into the current camera view; nearby is the most recent latent frame carrying full-resolution continuity. The prefix is processed by full self-attention with the target chunk, then sliced off. Because all four context streams are functions only of the past, per-chunk compute stays constant as roll-out gr

Load-bearing premise

The headline performance comparison assumes an even-handed benchmark protocol: the paper rewrites AlayaWorld's benchmark prompts into its training style but does not state whether the comparison models received the same treatment, and if they did not, the reported leads could reflect preprocessing rather than the model.

What would settle it

Run the benchmark twice, once with the same automated prompt adaptation applied to every model and once with none applied to any; if AlayaWorld's leads on brightness consistency, memory symmetry, or trajectory alignment shrink or invert under the even-handed protocol, the central 'best long-horizon performance' claim is settled against the model rather than for it.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Interactive roll-outs can in principle continue indefinitely: per-chunk compute is constant because the context is a fixed sink, a 6-frame history, and a capped cache, so the horizon is not limited by a growing context window.
  • Camera-controlled navigation and switchable text prompts can coexist in a single autoregressive model, so actions such as combat, spell-casting, or object appearance can be injected mid-roll-out without disturbing the existing scene.
  • The thirty-to-four step distillation retains the full control and memory stack, implying that few-step inference is not inherently at odds with long-horizon consistency.
  • Training on corrupted histories and self-generated residuals turns drift correction into a data-level property, implying that long-horizon stability can be improved without changing the architecture's capacity.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A natural extension the paper leaves implicit: because every context stream is a function of the past, the same prefix design could accept additional sensor streams—depth, semantics, object state—making the world model usable for closed-loop policy training inside the generated environment.
  • A cheap ablative test suggested by the design: truncate or disable the spatial-memory cache and re-measure memory-symmetry scores; if they collapse, the geometry-aligned cache is the cause of the reported memory gains rather than the temporal history.
  • The discrete distillation formulation, by avoiding Jacobian-vector products, may transfer to other chunked streaming diffusion models beyond interactive world models; this is an inference, not a claim in the paper.
  • Because the quantitative benchmark runs at 480p while the headline capability is 540p/720p, a direct test of whether the four-step stability and memory metrics carry to the higher resolutions is left open.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper presents AlayaWorld, a 15B video diffusion transformer that generates interactive long-horizon video by autoregressively producing short latent chunks under camera trajectories and switchable text prompts. The method combines a bounded visual context (persistent sink frame, compressed temporal history, geometry-aligned spatial memory, and recent-frame conditioning) with anti-drift training on corrupted histories and self-produced residuals, followed by a discrete autoregressive distillation that reduces inference to four sampling steps per chunk. The authors report state-of-the-art results on iWorld-Bench, including leading scores in brightness consistency, color-temperature constraint, sharpness retention, trajectory accuracy, memory symmetry, and trajectory alignment.

Significance. If the reported results hold, AlayaWorld demonstrates that a bounded-context autoregressive architecture with geometry-aligned spatial memory and drift-aware training can sustain stable, controllable, long-horizon interactive generation at low sampling cost. The paper is unusually detailed for a technical report: it describes the training-data mixture, curation pipeline, three-stage training procedure, and distillation objectives, and it ships code and a project page, which supports reproducibility and follow-up work. The quantitative iWorld-Bench comparison is the main evidence for the headline claim, so its fairness and statistical reliability are decisive.

major comments (2)
  1. [§4.2, Table 3] The benchmark comparison is potentially confounded by one-sided prompt adaptation. Section 4.2 states: 'Before inference, we apply an automated, semantics-preserving prompt adaptation procedure to reformulate the benchmark instructions into the prompt style used during training.' The paper does not state whether the seven comparison models (Cosmos, HunyuanVideo-1.5, WAN 2.2, YUME 1.5, Matrix-Game 2.0, HY-World 1.5) received the same reformulation. Since AlayaWorld was trained with prompts in exactly this style, whereas the baselines were trained with their own instruction formats, a one-sided adaptation would give AlayaWorld an in-distribution advantage on instruction-following metrics such as Trajectory Accuracy (0.7985 vs. next-best 0.7472) and Memory Symmetry (0.8871 vs. next-best 0.8481). This is a control gap, not evidence of intent, but it directly undercuts the 'best performance o
  2. [Table 3 and §4.2] Table 3 reports a single number per model per metric with no error bars, confidence intervals, number of evaluation episodes, or significance tests. Several decisive margins are small: Trajectory Alignment is 0.7018 vs. 0.6776, Memory Symmetry is 0.8871 vs. 0.8481, and Motion Smoothness is 0.9924 vs. 0.9921. Without uncertainty estimates or per-episode statistics, the claim that AlayaWorld 'achieves the best performance' on these metrics is not statistically supported. Please report variance across evaluation runs or seeds, or at least the number of benchmark clips and per-metric standard deviations; if the benchmark protocol supplies scores per episode, aggregate them transparently.
minor comments (4)
  1. [§4.1] The paper says AlayaWorld is evaluated on the 'standardized WorldMark test suite' with results 'publicly available at warena.ai', but no WorldMark results appear in the manuscript. Either include those results in a table or appendix, or remove the claim, since the current text invites verification without providing the evidence.
  2. [Abstract / Introduction] Grammar: 'It enable us to create...' should be 'This enables us...' in the abstract and introduction. Also, the phrase 'Conceived as a full-stack, open-source, and long-term project' appears in both the abstract and conclusion; consider varying it.
  3. [§3.4] The distillation objective in Eq. (7) writes the gradient of a KL divergence, but the right-hand side is an expectation of a product of score difference and \(\partial \hat z_i / \partial \theta\). This is a standard DMD gradient estimator, but the notation should be clarified to distinguish the loss from the update direction; as written, the equation is dimensionally ambiguous.
  4. [Table 3] The table lists models but does not state their inference settings (sampling steps, resolution, prompt template, number of generated chunks) for the comparison. A short protocol note, even in a footnote, would help readers assess whether the comparison is apples-to-apples beyond the prompt-adaptation issue.

Circularity Check

0 steps flagged

No circular derivation found; iWorld-Bench claim is external and self-contained. Minor self-referential evaluation elements (self-authored WorldMark citation, own Yume baseline) warrant a low score but do not reduce the claim.

full rationale

The headline claim is assessed on iWorld-Bench [5], an external benchmark, against external baselines in Table 3. The architecture components — sink frame, compressed temporal history, geometry-aligned spatial memory, anti-drift corrupted-history/error-bank training, and DMD + self-forcing++ + consistency distillation — are described with explicit objectives (Eqs. 4–8) trained on video data, not on the reported benchmark metrics. No fitted parameter is defined in terms of a quantity later renamed as a prediction; brightness consistency, trajectory accuracy, and memory metrics are measured by the benchmark. The prompt-adaptation procedure in Sec. 4.2 ('Before inference, we apply an automated, semantics-preserving prompt adaptation procedure to reformulate the benchmark instructions into the prompt style used during training') is a legitimate fairness/protocol concern for baseline comparison, but it is not a circular reduction of the model's derivation. Self-citations exist — Sekai-Real [17] and Yume [19,20] are co-authored by AlayaWorld members, and WorldMark [30] is an Alaya-authored benchmark invoked without reporting numbers — but the central result does not rest on these citations. Also, Sec. 3.1 invokes GEN3C and 'Frame Preservation' without references; this is a missing-support issue, not circularity. Overall, no step in the derivation chain reduces to its own input; score 2 reflects only minor self-referential evaluation elements.

Axiom & Free-Parameter Ledger

8 free parameters · 5 axioms · 0 invented entities

The paper introduces no new physical entities; its mechanisms (sink, temporal memory, spatial memory, error bank) are architectural modules assembled from cited components. The free-parameter list captures the hand-set hyperparameters and data-mixture choices that the reported performance depends on; the most consequential is the benchmark-side prompt adaptation, which is a tuned procedure rather than a fitted number.

free parameters (8)
  • per-axis camera-control scales = calibrated from real-data motion statistics (Sec. 3.3)
    AdaLN camera module scales are tuned to the training corpus's motion distribution; not derived from first principles.
  • Helios corruption strength rho = not specified; sample sigma~U(0,rho)
    Anti-drift data augmentation strength is a hand-chosen hyperparameter controlling how much context corruption the model sees.
  • occlusion tolerance delta = 0.1
    Spatial-memory z-buffer coverage tolerance set by hand; affects which past pixels are trusted during warping.
  • spatial-memory retrieval count = 10 frames
    Maximum number of past frames warped into the current view; a capacity/compute tradeoff chosen without analysis.
  • history length L = 6 latent frames
    Temporal-history window size; fixed without reported sensitivity analysis.
  • next-forcing loss weight = 0.5*L_nf
    Weight balancing the main flow-matching loss and the auxiliary next-forcing head (Eq. 6).
  • data mixture proportions = retuned per phase (Sec. 3.3)
    Per-source sampling weights are adjusted to balance domains; not reported numerically.
  • benchmark prompt adaptation = not specified
    Automated reformulation of iWorld-Bench instructions into AlayaWorld's caption style; a tuned preprocessing that affects the reported metrics.
axioms (5)
  • domain assumption LTX-2.3 public checkpoint supplies a strong general text-to-video prior that can be adapted to world modeling.
    Section 3 states AlayaWorld is built on LTX-2.3 (13B after removing audio); the entire system inherits this prior's capabilities and failure modes.
  • domain assumption Depth-Anything-3 monocular depth is accurate enough for cross-view warping in the spatial memory.
    Eq. (3) and Section 3.1 unproject and reproject frames using monocular depth; wrong depth injects geometrically inconsistent pixels into the context.
  • domain assumption ViPE recovers camera intrinsics and poses accurate enough for trajectory-conditioned training.
    Section 2.1 uses ViPE to recover poses for all real sources lacking camera metadata; any pose error propagates into the camera-control labels and spatial-memory cache.
  • domain assumption iWorld-Bench metrics are a valid proxy for interactive world-model quality.
    Section 4.1 adopts the iWorld-Bench protocol without questioning its metric validity; the headline claim is defined by this benchmark.
  • domain assumption Self-forcing++ scoring on self-generated rollouts reduces the train/inference gap of autoregressive generation.
    Section 3.4 asserts self-forcing++ is 'the key to seam-free chunk continuation' without an ablation isolating its effect.

pith-pipeline@v1.3.0-alltime-deepseek · 12516 in / 12424 out tokens · 106133 ms · 2026-08-01T15:43:51.845379+00:00 · methodology

0 comments
read the original abstract

Unlike conventional video game development, which relies on labor-intensive pipelines for asset production, animation, physics, and programming, video world models generate interactive environments from user inputs instantly. It enable us to create customized, explorable, and continuously evolving virtual world from text, an image, or video. Realizing this vision requires four tightly coupled capabilities: interaction, persistent spatiotemporal consistency, stable long-horizon generation, and efficient response. We present AlayaWorld, an interactive long-horizon video world model that generates 24-fps video at 540p and 720p. Built on a 15B video diffusion transformer, AlayaWorld generates short latent chunks autoregressively under camera trajectories and switchable text prompts. Its bounded visual context combines a persistent sink frame, compressed temporal history, geometry-aligned spatial memory, and recent-frame conditioning. To reduce long-term drift, the model is trained with corrupted histories and prediction residuals collected from its own roll-outs. We further introduce a discrete autoregressive distillation formulation that combines distribution-matching distillation, self-forcing++, and consistency distillation, reducing inference from approximately 30 sampling steps to four steps per chunk. On iWorld-Bench, AlayaWorld achieves the best performance over long-horizon generation. Conceived as a full-stack, open-source, and long-term project, AlayaWorld is intended to provide an extensible foundation for future research on interactive video world models.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

34 extracted references · 16 linked inside Pith

  1. [1]

    Agarwal, A

    N. Agarwal, A. Ali, M. Bala, Y. Balaji, E. Barker, T. Cai, P . Chattopadhyay, Y. Chen, Y. Cui, Y. Ding, et al. Cosmos world foundation model platform for physical ai.arXiv preprint arXiv:2501.03575, 2025

    Pith/arXiv arXiv 2025

  2. [2]

    Assran, A

    M. Assran, A. Bardes, D. Fan, Q. Garrido, R. Howes, M. Komeili, M. Muckley, A. Rizvi, C. Roberts, K. Sinha, et al. V-jepa 2: Self-supervised video models enable understanding, prediction and planning. arXiv preprint arXiv:2506.09985, 2025

    Pith/arXiv arXiv 2025

  3. [3]

    Comanici, E

    G. Comanici, E. Bieber, M. Schaekermann, I. Pasupat, N. Sachdeva, I. Dhillon, M. Blistein, O. Ram, D. Zhang, E. Rosen, et al. Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and next generation agentic capabilities.arXiv preprint arXiv:2507.06261, 2025

    Pith/arXiv arXiv 2025

  4. [4]

    J. Cui, J. Wu, M. Li, T. Yang, X. Li, R. Wang, A. Bai, Y. Ban, and C.-J. Hsieh. Self-forcing++: Towards minute-scale high-quality video generation.arXiv preprint arXiv:2510.02283, 2025

    Pith/arXiv arXiv 2025

  5. [5]

    J. Fang, Y. Lei, Q. Wan, Z. Wang, Y. Huang, Y. Xu, B. Zhao, W. Zhang, C. Gao, X. Chen, et al. iworld- bench: A benchmark for interactive world models with a unified action generation framework.arXiv preprint arXiv:2605.03941, 2026

    Pith/arXiv arXiv 2026

  6. [6]

    Gemma 3 technical report.arXiv preprint arXiv:2503.19786, 2025

    Gemma Team. Gemma 3 technical report.arXiv preprint arXiv:2503.19786, 2025

    Pith/arXiv arXiv 2025

  7. [7]

    HaCohen, N

    Y. HaCohen, N. Chiprut, B. Brazowski, D. Shalem, D. Moshe, E. Richardson, E. Levin, G. Shiran, N. Zabari, O. Gordon, et al. Ltx-video: Realtime video latent diffusion.arXiv preprint arXiv:2501.00103, 2024

    Pith/arXiv arXiv 2024

  8. [8]

    C. He, Q. Zheng, R. Zhu, X. Zeng, Y. Fan, and Z. Tu. Cover: A comprehensive video quality evaluator. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, pages 5799–5809, 2024

    2024

  9. [9]

    H. He, C. Yang, S. Lin, Y. Xu, M. Wei, L. Gui, Q. Zhao, G. Wetzstein, L. Jiang, and H. Li. Cameractrl ii: Dynamic scene exploration via camera-controlled video diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 13416–13426, 2025

    2025

  10. [10]

    X. He, C. Peng, Z. Liu, B. Wang, Y. Zhang, Q. Cui, F. Kang, B. Jiang, M. An, Y. Ren, et al. Matrix-game 2.0: An open-source real-time and streaming interactive world model.arXiv preprint arXiv:2508.13009, 2025. 14

    Pith/arXiv arXiv 2025

  11. [11]

    Huang, Q

    J. Huang, Q. Zhou, H. Rabeti, A. Korovko, H. Ling, X. Ren, T. Shen, J. Gao, D. Slepichev, C.-H. Lin, J. Ren, K. Xie, J. Biswas, L. Leal-Taixe, and S. Fidler. Vipe: Video pose engine for 3d geometric perception.arXiv preprint arXiv:2508.10934, 2025

    Pith/arXiv arXiv 2025

  12. [12]

    Huang, Y

    Z. Huang, Y. He, J. Yu, F. Zhang, C. Si, Y. Jiang, Y. Zhang, T. Wu, Q. Jin, N. Chanpaisit, et al. Vbench: Comprehensive benchmark suite for video generative models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 21807–21818, 2024

    2024

  13. [13]

    Easyocr: Ready-to-use ocr with 80+ supported languages

    Jaided AI. Easyocr: Ready-to-use ocr with 80+ supported languages. https://github.com/JaidedAI/ EasyOCR, 2020

    2020

  14. [14]

    Khanam and M

    R. Khanam and M. Hussain. Yolov11: An overview of the key architectural enhancements.arXiv preprint arXiv:2410.17725, 2024

    Pith/arXiv arXiv 2024

  15. [15]

    Kimi k2.6.https://huggingface.co/moonshotai/Kimi-K2.6, 2026

    Kimi Team. Kimi k2.6.https://huggingface.co/moonshotai/Kimi-K2.6, 2026

    2026

  16. [16]

    W. Li, W. Pan, P .-C. Luan, Y. Gao, and A. Alahi. Stable video infinity: Infinite-length video generation with error recycling.arXiv preprint arXiv:2510.09212, 2025

    arXiv 2025

  17. [17]

    Z. Li, C. Li, X. Mao, S. Lin, M. Li, S. Zhao, Z. Xu, X. Li, Y. Feng, J. Sun, Z. Li, F. Zhang, J. Ai, Z. Wang, Y. Wu, T. He, J. Pang, Y. Qiao, Y. Jia, and K. Zhang. Sekai: A video dataset towards world exploration. arXiv preprint arXiv:2506.15675, 2025

    arXiv 2025

  18. [18]

    L. Ling, Y. Sheng, Z. Tu, W. Zhao, C. Xin, K. Wan, L. Yu, Q. Guo, Z. Yu, Y. Lu, et al. Dl3dv-10k: A large-scale scene dataset for deep learning-based 3d vision. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 22160–22169, 2024

    2024

  19. [19]

    X. Mao, Z. Li, C. Li, X. Xu, K. Ying, and K. Zhang. Yume1.5: A text-controlled interactive world generation model. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 7752–7761, 2026

    2026

  20. [20]

    X. Mao, S. Lin, Z. Li, C. Li, W. Peng, T. He, J. Pang, M. Chi, Y. Qiao, and K. Zhang. Yume: An interactive world generation model.arXiv preprint arXiv:2507.17744, 2025

    Pith/arXiv arXiv 2025

  21. [21]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P . Mishkin, J. Clark, et al. Learning transferable visual models from natural language supervision. InInternational Conference on Machine Learning (ICML), pages 8748–8763, 2021

    2021

  22. [22]

    Y. Song, P . Dhariwal, M. Chen, and I. Sutskever. Consistency models. InInternational Conference on Machine Learning (ICML), 2023

    2023

  23. [23]

    Hy-world 1.5: A systematic framework for interactive world modeling with real-time latency and geometric consistency.arXiv preprint, 2025

    Team HunyuanWorld. Hy-world 1.5: A systematic framework for interactive world modeling with real-time latency and geometric consistency.arXiv preprint, 2025

    2025

  24. [24]

    Teed and J

    Z. Teed and J. Deng. Raft: Recurrent all-pairs field transforms for optical flow. InProceedings of the European Conference on Computer Vision (ECCV), pages 402–419, 2020

    2020

  25. [25]

    Hunyuanvideo 1.5 technical report, 2025

    Tencent Hunyuan Foundation Model Team. Hunyuanvideo 1.5 technical report, 2025

    2025

  26. [26]

    Tschannen, A

    M. Tschannen, A. Gritsenko, X. Wang, M. F. Naeem, I. Alabdulmohsin, N. Parthasarathy, T. Evans, L. Beyer, Y. Xia, B. Mustafa, et al. Siglip 2: Multilingual vision-language encoders with improved semantic understanding, localization, and dense features.arXiv preprint arXiv:2502.14786, 2025

    Pith/arXiv arXiv 2025

  27. [27]

    B. Wang, G. Xu, Z. Tang, J. Zhang, and Z. Cheng. Omnishotcut: Holistic relational shot boundary detection with shot-query transformer.arXiv preprint arXiv:2604.24762, 2026

    Pith/arXiv arXiv 2026

  28. [28]

    J. Wang, Y. Yuan, R. Zheng, Y. Lin, J. Gao, L.-Z. Chen, Y. Bao, Y. Zhang, C. Zeng, Y. Zhou, et al. Spatialvid: A large-scale video dataset with spatial annotations.arXiv preprint arXiv:2509.09676, 2025

    arXiv 2025

  29. [29]

    G. Xu, Q. Zhang, J. Zhou, X. Zhu, Y. Shen, X. Yang, and Y. Xu. Next forcing: Causal world modeling with multi-chunk prediction.arXiv preprint arXiv:2606.11187, 2026

    Pith/arXiv arXiv 2026

  30. [30]

    X. Xu, Z. Lin, K. He, Y. Feng, X. Mao, Y. Yin, K. Zhang, and Y. Ge. Worldmark: A unified benchmark suite for interactive video world models.arXiv preprint arXiv:2604.21686, 2026. 15

    Pith/arXiv arXiv 2026

  31. [31]

    T. Yin, M. Gharbi, T. Park, R. Zhang, E. Shechtman, F. Durand, and W. T. Freeman. Improved distribution matching distillation for fast image synthesis.Advances in neural information processing systems, 37:47455–47487, 2024

    2024

  32. [32]

    S. Yuan, Y. Yin, Z. Li, X. Huang, X. Yang, and L. Yuan. Helios: Real real-time long video generation model.arXiv preprint arXiv:2603.04379, 2026

    arXiv 2026

  33. [33]

    Zheng, G

    K. Zheng, G. He, M. Zhao, J. Zhang, H. Chen, J. Chen, C.-H. Lin, M.-Y. Liu, J. Zhu, and Q. Ma. Causal- rcm: A unified teacher-forcing and self-forcing open recipe for autoregressive diffusion distillation in streaming video generation and interactive world models.arXiv preprint arXiv:2606.25473, 2026

    Pith/arXiv arXiv 2026

  34. [34]

    T. Zhou, R. Tucker, J. Flynn, G. Fyffe, and N. Snavely. Stereo magnification: Learning view synthesis using multiplane images.ACM Transactions on Graphics (SIGGRAPH), 37(4):1–12, 2018. 16

    2018