Pith. sign in

REVIEW 4 major objections 5 minor 21 cited by

Fast Video Generation with Sliding Tile Attention

T0 review · 4 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read Sliding tile attention replaces wasteful full 3D attention in video diffusion models, cutting generation time nearly in half without retraining while keeping output quality at near-parity.

desk verdict A genuinely useful sparse-attention kernel for video DiTs with real speedups, but the paper's own quality numbers and latency accounting undercut its headline claims. read the letter →

arxiv 2502.04507 v3 pith:K4GPZFZU submitted 2025-02-06 cs.CV

Peiyuan Zhang , Yongqi Chen , Runlong Su , Hangliang Ding , Ion Stoica , Zhengzhong Liu , Hao Zhang This is my paper · ORCID
classification cs.CV
keywords slidingtileattentionvideodiffusiontransformersparse3DlocalityheadspecializationkerneloptimizationHunyuaninferenceacceleration
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that the dominant compute cost of state-of-the-art video diffusion transformers is full 3D attention, and that most of this attention is redundant: a local window covering about 15.5% of the token space carries about 70% of the attention mass. To exploit this, it introduces sliding tile attention (STA), which slides over cubes of tokens rather than individual tokens, so every computed attention block is dense and the sparse mask never touches the fast compute path. On a leading open video model, STA cuts end-to-end generation of a 5-second 720P clip from 945 seconds with FlashAttention-3 to 501 seconds with no training, with an 83% human tie rate against the original, and to 268 seconds after fine-tuning with a VBench Total of 82.62 versus 82.71. If right, this gives a plug-in speedup for existing video diffusion transformers and a new system-algorithm co-design pattern for sparse attention.

What carries the argument

The load-bearing object is the tile: a contiguous group of tokens forming a spatial-temporal cube, sized so that one tile equals one FlashAttention block, with token ordering inside a tile kept consecutive. STA slides over tiles instead of tokens, so all queries in a tile share the same key-tile set and the attention map contains only dense and empty blocks. The kernel's efficiency comes from a consumer–producer split: producer warpgroups asynchronously load only the KV blocks in the window and own all mask logic, while consumer warpgroups run dense attention with no masking overhead at all. The second mechanism is head specialization, exploited by Algorithm 1, which picks per-head window sizes by profiling 16 prompts to find the mask minimizing MSE to full attention.

What would settle it

Run the training-free STA configuration on a held-out set of prompts with strong camera motion, rapid cuts, or small fast-moving objects, and compare against full attention with human evaluation; if the tie rate falls far below the reported 83% or VBench Total drops more than a point, the head-specialization assumption does not transfer. A kernel-level check: measure the STA kernel's MFU and latency on a non-Hopper GPU; the paper's hardware claim predicts the wall-clock speedup still scales with sparsity, so a machine where this fails would falsify the generality of the tile-based design.

Watch

Extended reading notes

Core claim

The central discovery is that video diffusion transformers trained with full 3D attention develop strong, prompt-invariant locality: each attention head concentrates mass in a fixed local spatial–temporal window, and different heads specialize in different window sizes. The paper shows that by reordering the attention computation into tiles whose size equals the FlashAttention block size, the sliding-window pattern decomposes into dense blocks and empty blocks only, with no mixed blocks, turning a theoretically cheap but practically slow sparse attention into one whose wall-clock speedup tracks the FLOP reduction. Per-head window sizes can then be selected automatically by minimizing the MSE between masked and full attention outputs on just 16 prompts, giving training-free speedups with near-parity quality, and a short fine-tuning stage closes the remaining gap.

Load-bearing premise

The training-free speedup rests on the assumption that the per-head window sizes chosen by minimizing MSE on only 16 prompts generalize to all other prompts, rather than overfitting those sixteen examples.

Editorial extensions

If this is right

  • Existing pretrained video diffusion transformers can be accelerated about 1.9x without any training by swapping in STA and a per-head window configuration.
  • With roughly 8 hours of fine-tuning on 2,000 generated clips, the same models run about 3.5x faster at 91% attention sparsity with essentially unchanged VBench scores.
  • STA outperforms cache-based acceleration on both quality and speed, including higher human win rates at comparable inference budgets.
  • Because STA's windows are selected per head, the method is complementary to step-distillation and caching, so the speedups compose.
  • The tile-based pattern also yields speedups in 2D image super-resolution, indicating the mechanism generalizes beyond video.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The prompt-invariance of head locality, if it holds, suggests that full 3D attention in video diffusion transformers is heavily overtrained, and the 16-prompt search could itself be replaced by an online adaptive search that updates windows per diffusion timestep without hurting quality.
  • The tile-wise decomposition may transfer to other high-dimensional attention problems with locality, such as point cloud transformers or multi-view 3D generators, where mixed-block overhead has similarly prevented sparse attention from paying off.
  • The reported quality numbers imply a direct cost-quality knob: because larger windows preserve more of the full-attention distribution, deployment could tune sparsity per scene complexity rather than globally.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes Sliding Tile Attention (STA), a tile-based sparse attention mechanism for video diffusion transformers. STA groups tokens into 3D tiles so that all queries in a tile attend to the same set of key tiles, turning the irregular sliding-window mask into a mix of dense and empty FlashAttention blocks. The authors implement STA with a ThunderKittens/FlashAttention-3-style kernel using asynchronous data loading and report large kernel speedups (up to 10.45x over FA3 at 91% sparsity, 58.79% MFU). On HunyuanVideo they report end-to-end latency reductions from 945s (FA3) to 501s training-free and 268s with fine-tuning, and quality evaluations via VBench, SSIM/PSNR/CD-FVD, and pairwise human evaluation. They also apply STA to FLUX for image super-resolution.

Significance. The kernel contribution is significant. The paper identifies a real gap: existing 2D/3D sliding-window attention implementations (NATTEN, CLEAR) fail to convert FLOP reductions into wall-clock speedups, and STA addresses this with a clean tile-based formulation and a carefully engineered kernel. The benchmarks are detailed and the code is promised to be public. If the measurement-scope and quality-claim issues are corrected, this would be a useful systems contribution for accelerating video diffusion models. The paper also gives credit for the disclosed mask-search design choice and provides a transparent kernel-efficiency table.

major comments (4)
  1. [Section 4 and Abstract] The end-to-end latency comparison is ambiguous in scope. Section 4 states, 'For end-to-end speedup on DiT, we report measured wall-clock latency, excluding time spent on VAE and text encoder,' while the abstract and Figure 1 describe 945s as 'total inference time' with attention taking 800s of it. If 945s is total pipeline time and the STA timings of 501s and 268s exclude VAE and text encoder, then the reported speedups of 1.89x and 3.53x compare different measurement scopes and are inflated. If 945s is intended to be DiT-only, then the abstract's wording is incorrect, and for any added VAE+text-encoder time E the user-facing speedups become (945+E)/(501+E) and (945+E)/(268+E), which are strictly smaller than the headline values. The paper should either report full-pipeline latencies including VAE and text encoder or explicitly state the value of E and recompute the speedups.
  2. [Abstract and Section 4.2 / Table 4] The claim of 'without quality degradation' is contradicted by the paper's own data. The abstract states that STA reduces latency to 501s 'without quality degradation,' but Table 4 reports VBench Total 80.58 for the training-free STA at 91% sparsity versus 82.71 for FA3, and Section 4.2 reports that STA-tf-1.89x has a 7.0-percentage-point lower win rate than loss rate in the human pairwise evaluation. These results indicate measurable quality loss, even if modest. The abstract should be softened to 'minimal quality degradation' or accompanied by a statistical test showing the human-eval deficit is not significant. Additionally, the two abstracts in the manuscript disagree on the training-free latency: the arXiv-style abstract says 685s while the paper's abstract says 501s; this inconsistency must be resolved.
  3. [Theorem 3.1 and Table 1] The formulas in Theorem 3.1 do not reproduce the numbers in Table 1. For the stated configuration (tile size 4, window size 11, video size 48), the theorem gives Ndense = (max(2*floor(12/8)-1,0))^3 * 12^3 = 1728 and Nmix = (2*floor(10/8)+1)^3 * 12^3 - 1728 = 44928. Since the total number of FA blocks is 12^3 * 12^3 = 2,985,984, these correspond to 0.06% dense and 1.50% mixed, but Table 1 reports 7.17% mixed. The text says 'We apply Theorem 3.1 and Theorem 3.2 to calculate the ratio of different blocks and report them in Table 1,' so this is an internal inconsistency. If boundary effects explain the difference, the theorem should state that it ignores boundaries and the table should be annotated accordingly. The qualitative claim that STA eliminates mixed blocks is not in question, but the quantitative comparison is inaccurate as presented.
  4. [Algorithm 1 and Section 3.2] The training-free configuration search in Algorithm 1 selects per-head window sizes by minimizing MSE to full-attention outputs on only 16 prompts. The claim that this configuration transfers to unseen prompts is central to the training-free quality results. The paper should provide a sensitivity analysis, for example by running the search on different random subsets of 16 prompts and reporting the resulting VBench scores or the variation in the selected window-size distribution. Without such an analysis, the VBench and human-eval results could be dependent on the particular 16-prompt search set, and the generality of the head-specialization assumption is not fully established.
minor comments (5)
  1. [Figure 3 and surrounding text] The caption and text contain corrupted unicode sequences such as '/uni00000013/uni00000018' and 'uni0000002b/uni00000048'; the figure labels should be regenerated with a proper font/encoding.
  2. [Section 1] The phrase 'head specialization remains agonistic to prompts' should read 'agnostic'.
  3. [Section 1] The phrase 'wallcock speedups' should read 'wall-clock speedups'.
  4. [Appendix A] The heading 'Visialization of 2D SWA' should read 'Visualization of 2D SWA'.
  5. [Section 4.4, footnote 2] The paper notes that 'Other memory-bound operations, such as LayerNorm and modulation, likely contribute to inference overhead,' but does not quantify this. A brief breakdown of the DiT inference time would help readers reconcile the 5.76x FLOP reduction with the 3.53x latency reduction.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: STA's speed and quality claims rest on independent kernel benchmarks and external quality evaluations; the one overlapping self-citation is not load-bearing.

full rationale

The derivation chain is self-contained against external evidence. The locality premise is measured directly from HunyuanVideo attention maps (Figures 2–3), not assumed from the method. The tiling construction and Theorems 3.1–3.2 are counting arguments based on explicit definitions, not fitted predictions. The per-head window search (Algorithm 1) is a disclosed profiling heuristic minimizing MSE to full-attention outputs; the resulting configuration is then evaluated on VBench and human preference, which are external to the search objective. The finetuning objective includes an attention-distillation term matching the teacher, but the paper's quality claim is subsequently checked on VBench rather than asserted by construction. Speedup claims are benchmarked against FA2, FA3, NATTEN, CLEAR, and Swin in Table 2, and quality is benchmarked against ∆-DiT and the original model. The only noteworthy issue is a latency-scope inconsistency: the abstract calls 945s "total inference time" while §4 reports measured wall-clock latency "excluding time spent on VAE and text encoder" for the reported end-to-end speedups. This affects the comparability and magnitude of the headline speedup but is a measurement-scope concern, not circularity. One citation (Li et al. 2024a) overlaps with the authors and is used in the sentence about keeping full attention for initial timesteps, but the same practice is also supported by two external citations and is not load-bearing. Overall, no circular step reduces a prediction to a fitted input or to a self-citation chain.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The STA method depends on three empirical premises: locality of attention in video DiTs, cross-prompt stability of head specialization, and the representativeness of HunyuanVideo plus the chosen quality benchmarks. The free parameters are the per-head window sizes (searched on 16 prompts), the full-attention warmup schedule, and the fine-tuning loss weights. These are disclosed hyperparameters, but they are not derived from first principles and they materially affect the reported trade-off.

free parameters (3)
  • Per-head window size configuration = Not tabulated; examples w=(18,24,24), w=(30,40,40), w=(30,24,40)
    Selected by Algorithm 1 to minimize MSE to full-attention outputs on 16 prompts; the choice directly sets sparsity and the reported quality/speedup trade-off.
  • Full-attention warmup steps T0 = 12 (50-step), 6 (25-step), 3 (10-step)
    Hand-chosen following prior caching methods; it controls how much expensive full attention is kept and therefore affects both speedup and quality.
  • Finetuning loss coefficients (alpha, beta, gamma) = 1, 0.5, 0.5
    Set in Appendix B without sensitivity analysis; the balance between attention distillation, final-output loss, and data loss affects trained quality.
assumptions (4)
  • domain assumption Attention scores in pretrained video DiTs concentrate in a local 3D window
    Validated on HunyuanVideo (Figures 2, 3); if locality breaks, STA's window will discard needed context. Section 1.
  • domain assumption Head specialization is prompt-agnostic
    Used to transfer per-head windows found on 16 prompts to all evaluation prompts. Section 3.2 and Figure 3.
  • domain assumption HunyuanVideo is representative of state-of-the-art video DiTs
    Main end-to-end evaluation is on HunyuanVideo only; other open models are skipped due to lower quality (Section 4). Generalization is assumed.
  • domain assumption VBench and human pairwise preference measure video quality adequately
    Used to support 'no quality degradation'; these metrics may not capture all failure modes.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fast Video Generation with Sliding Tile Attention." pith.science (2026). https://pith.science/paper/K4GPZFZU

@misc{pith2026250204507,
  author       = {Pith},
  title        = {Pith review of: Fast Video Generation with Sliding Tile Attention},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/K4GPZFZU}},
  note         = {Machine review of arXiv:2502.04507}
}
read the original abstract

Diffusion Transformers (DiTs) with 3D full attention power state-of-the-art video generation, but suffer from prohibitive compute cost -- when generating just a 5-second 720P video, attention alone takes 800 out of 945 seconds of total inference time. This paper introduces sliding tile attention (STA) to address this challenge. STA leverages the observation that attention scores in pretrained video diffusion models predominantly concentrate within localized 3D windows. By sliding and attending over the local spatial-temporal region, STA eliminates redundancy from full attention. Unlike traditional token-wise sliding window attention (SWA), STA operates tile-by-tile with a novel hardware-aware sliding window design, preserving expressiveness while being hardware-efficient. With careful kernel-level optimizations, STA offers the first efficient 2D/3D sliding-window-like attention implementation, achieving 58.79% MFU. Precisely, STA accelerates attention by 2.8-17x over FlashAttention-2 (FA2) and 1.6-10x over FlashAttention-3 (FA3). On the leading video DiT, HunyuanVideo, STA reduces end-to-end latency from 945s (FA3) to 685s without quality degradation, requiring no training. Enabling finetuning further lowers latency to 268s with only a 0.09% drop on VBench. We make our codebase public at https://github.com/hao-ai-lab/FastVideo.

Figures

Figures reproduced from arXiv: 2502.04507 by the authors.

Figure 1
Figure 1. (a) Generating a 5s 720P clip in Hunyuan involves pro￾cessing 115K tokens, making attention the dominant cost. (b) Attention latency comparison: existing methods fail to trans￾late FLOP reduction into wall-clock speedup; STA is hardware￾efficient and achieves proportional speedup with sparsity. synthesizing long-duration, visually coherent outputs (Pee￾bles & Xie, 2023; OpenAI, 2024). Central to their success is 3D … view at source ↗
Figure 2
Figure 2. Visualization of attention locality. The green point means the query point and the magma-colored regions indicate areas of high attention values in response to the query. Instead of attending to the entire image, the query’s attention forms a concentrated local hotspot. attention scores concentrated within a local window. As shown in [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Left: Fraction of attention scores within a (12, 24, 24) local window across diffusion steps and 10 different prompts. Most heads show high recall, indicating a local attention pattern. Right: Despite the different recall across heads, the standard deviation across prompts remains low. This observation seemingly suggests that sliding window attention (SWA) is an ideal alternative to retain attention expressiveness w… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: The attention map of NATTEN, Tiled NATTEN, and STA. We plot with an image size 24×24 and a 12×12 local window. The tile size is set to 4×4. (a) NATTEN creates many mixed blocks that are very inefficient for Flash Attention computation. (b) Tiled NATTEN increases the nu…
Figure 5
Figure 5. Figure 5: 2D SLIDING TILE ATTENTION with tile size (2, 2) and window size (6, 6). After attending to all the key tiles, each query tile will generate nine 4x4 dense blocks in the attention map. We showcase 2D STA for better illustration. 3D STA can be inferred similarly. mixed b…
Figure 6
Figure 6. Figure 6: Qualitative example of 720P 5-second videos. While fine-tuning introduces minor shifts in the output distribution of STA-t-2.43x, the model still preserves high video generation quality. Videos generated by ∆-DiT are generally less sharp than those generated by the ori…
Figure 7
Figure 7. Figure 7: Human evaluation on 200 prompts from the MovieGen Bench (Polyak et al., 2024). STA achieves a 1.89× end-to-end speedup while maintaining performance comparable to the original HunyuanVideo. Additionally, STA consistently outperforms ∆- DiT across different inference bu…
Figure 8
Figure 8. Figure 8: Left: Conventional zigzag flattening strategy. Right: STA’ sequence flattening strategy. The plot is given assuming a (9, 9) image with (3, 3) tile size [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: 2D Sliding Window Attention visualization. and better information exchange. Swin attention is typically used in a train-from-scratch setting. A limitation of this approach is that it disrupts local connectivity within a single attention layer. Tokens in adjacent region…
Figure 10
Figure 10. Figure 10: Qualitative comparisons. While fine-tuning introduces minor shifts in the output distribution of STA-t-2.43x, the model still preserves high video generation quality. Videos generated by ∆-DiT are generally less sharp than those generated by the original HunyuanVideo …
Figure 11
Figure 11. Figure 11: Qualitative comparisons. While fine-tuning introduces minor shifts in the output distribution of STA-t-2.43x, the model still preserves high video generation quality. Videos generated by ∆-DiT are generally less sharp than those generated by the original HunyuanVideo …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 21 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Dynamic-in-Few-Step: Unifying Dynamic Computation and Few-Step Distillation for Efficient Video Generation

    cs.CV 2026-07 conditional novelty 6.5 of 10

    Joint few-step distillation and step-specific structural pruning turns a video diffusion model into a compact Mixture-of-Models that cuts 24% extra FLOPs per step and reaches 30× speedup on Wan-14B.

  2. RotateAttention: RoPE-Aware Rotation and Range Rectification for INT4 Quantized Attention in Video Generation

    cs.CV 2026-07 conditional novelty 6.5 of 10

    RoPE-aware orthogonal rotations plus full-range INT4 quantization of attention probabilities deliver mixed-precision FlashAttention that matches FP16 video quality at 1.5–1.68× end-to-end speedup.

  3. Context-Matched Distillation: Teacher Causality for Autoregressive Video Distillation

    cs.CV 2026-08 conditional novelty 6.0 of 10

    A causal-distillation method that scores each generated video frame only under its causal history improves few-step autoregressive video quality and camera-control adherence.

  4. SparSTAR: Sparse Attention for SpaceTime AutoRegressive Video Synthesis

    cs.CV 2026-08 conditional novelty 6.0 of 10

    SparSTAR dynamically picks which key blocks each attention head reads at expensive video scales, yielding about a 1.6x end-to-end speedup while VBench scores stay within 0.1 points of dense InfinityStar.

  5. SAF3R: Dynamic Sparse Attention for Feed-Forward 3D Reconstruction Transformers

    cs.CV 2026-07 conditional novelty 6.0 of 10

    Training-free head-wise dynamic sparse attention preserves F3R pose and reconstruction quality while delivering multi-fold end-to-end speedups on long image sequences.

  6. SnapGen++: Unleashing Diffusion Transformers for Efficient High-Fidelity Image Generation on Edge Devices

    cs.CV 2026-01 conditional novelty 6.0 of 10

    A compact elastic diffusion transformer with adaptive sparse attention and knowledge-guided distribution-matching distillation achieves 4-step 1K image generation on a phone in roughly 1.8 seconds.

  7. FreeLong++: Training-Free Long Video Generation via Multi-band SpectralFusion

    cs.CV 2025-06 conditional novelty 6.0 of 10

    FreeLong++ extends short-video diffusion models to 4x to 8x longer clips, without retraining, by fusing multiple windowed attention branches through frequency-domain filters and a spectral noise initialization.

  8. VMoBA: Mixture-of-Block Attention for Video Diffusion Models

    cs.CV 2025-06 conditional novelty 6.0 of 10

    VMoBA is a sparse attention mechanism for video diffusion models that combines cyclic 1D-2D-3D block partitioning with global and threshold-based block selection to reduce training FLOPs while keeping generation quality.

  9. FPSAttention: Training-Aware FP8 and Sparsity Co-Design for Fast Video Diffusion

    cs.CV 2025-06 conditional novelty 6.0 of 10

    FPSAttention co-designs FP8 quantization and sparsity with training, achieving 4.96x end-to-end video generation speedup on Wan2.1 with roughly preserved quality.

  10. Chipmunk: Training-Free Acceleration of Diffusion Transformers with Dynamic Column-Sparse Deltas

    cs.CV 2025-06 conditional novelty 6.0 of 10

    Chipmunk speeds up diffusion transformer inference by recomputing, at each step, only the top percent of attention and MLP activation columns that change most between steps, caching the rest in column-sparse GPU kernels.

  11. Dual-Expert Consistency Model for Efficient and High-Quality Video Generation

    cs.CV 2025-06 conditional novelty 6.0 of 10

    By training a semantic expert and a LoRA-based detail expert, DCM reaches nearly teacher-level VBench scores with 4-step video sampling on HunyuanVideo and CogVideoX.

  12. Grouping First, Attending Smartly: Training-Free Acceleration for Diffusion Transformers

    cs.CV 2025-05 conditional novelty 6.0 of 10

    GRAT accelerates pretrained diffusion transformers by grouping tokens and restricting each group's attention to neighboring blocks or criss-cross rows and columns, achieving large speedups with near-full-attention quality.

  13. Generalized Neighborhood Attention: Multi-dimensional Sparse Attention at the Speed of Light

    cs.CV 2025-04 conditional novelty 6.0 of 10

    Generalized Neighborhood Attention adds a stride parameter to neighborhood attention, aligning windows with GPU tiles so sparse attention achieves its full theoretical speedup; a Blackwell kernel delivers 1.26x-1.63x ...

  14. MMInference: Accelerating Pre-filling for Long-Context VLMs via Modality-Aware Permutation Sparse Attention

    cs.CV 2025-04 conditional novelty 6.0 of 10

    MMInference speeds up long-context VLM prefill by up to 8.3x at 1M tokens using modality-aware permutation sparse attention while keeping accuracy close to full attention.

  15. SPADE: An Input-Adaptive Sparse Attention Engine for Fast Video Diffusion Models Inference

    cs.CV 2026-08 conditional novelty 5.0 of 10

    SPADE combines static, semi-static, and dynamic block-sparse attention with a cheap SICS-based blocking heuristic to speed up video diffusion inference by up to 1.80x end-to-end.

  16. ENA: Efficient N-dimensional Attention

    cs.LG 2025-08 conditional novelty 5.0 of 10

    ENA combines linear recurrence with hardware-friendly sliding tile attention to model images and videos efficiently, claiming Transformer-level accuracy at roughly 70% attention sparsity.

  17. From Black Box to Transparency: Enhancing Automated Interpreting Assessment with Explainable AI in College Classrooms

    cs.CL 2025-08 unverdicted novelty 5.0 of 10

    An explainable model using BLEURT, CometKiwi, pause features, and Chinese phraseological diversity predicts human-rated quality dimensions in English-Chinese consecutive interpreting, with SHAP identifying the stronge...

  18. Hierarchical Fine-grained Preference Optimization for Physically Plausible Video Generation

    cs.CV 2025-08 conditional novelty 4.0 of 10

    A hierarchical direct preference optimization with four alignment levels plus automated data selection improves physical plausibility of text-to-video models.

  19. Critique of World Model

    cs.LG 2025-07 conditional novelty 4.0 of 10

    The paper argues world models should simulate actionable possibilities and proposes GLP, a hierarchical generative architecture that closes the loop with observation reconstruction, but provides no experiments.

  20. PAROAttention: Pattern-Aware ReOrdering for Efficient Sparse and Quantized Attention in Visual Generation Models

    cs.CV 2025-06 conditional novelty 4.0 of 10

    PAROAttention permutes tokens along frame, height, and width axes to make visual attention block-wise, enabling sparse and INT8/INT4 quantized attention with near-baseline generation quality.

  21. Reinforcement Learning: From Algorithms To Foundation Models

    cs.AI 2026-07 conditional novelty 3.0 of 10

    A dissertation uniting the author's published results: non-exploitable Nash-DQN policies and the FightLadder benchmark for games, plus diffusion/consistency-model world models for RL — a compilation rather than new results.

Reference graph

Works this paper leans on

46 extracted references · 5 canonical work pages · cited by 21 Pith papers

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    Arar, M., Shamir, A., and Bermano, A. H. Learned queries for efficient local attention. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 10841--10852, 2022

    work page 2022

  3. [3]

    E., and Cohan, A

    Beltagy, I., Peters, M. E., and Cohan, A. Longformer: The long-document transformer, 2020. URL https://arxiv.org/abs/2004.05150

    arXiv 2020

  4. [4]

    Black-Forest. Flux. https://github.com/black-forest-labs/flux, 2023

    work page 2023

  5. [5]

    Videocrafter1: Open diffusion models for high-quality video generation, 2023

    Chen, H., Xia, M., He, Y., Zhang, Y., Cun, X., Yang, S., Xing, J., Liu, Y., Chen, Q., Wang, X., Weng, C., and Shan, Y. Videocrafter1: Open diffusion models for high-quality video generation, 2023. URL https://arxiv.org/abs/2310.19512

    arXiv 2023

  6. [6]

    Delta-dit: A training-free acceleration method tailored for diffusion transformers

    Chen, P., Shen, M., Ye, P., Cao, J., Tu, C., Bouganis, C.-S., Zhao, Y., and Chen, T. Delta-dit: A training-free acceleration method tailored for diffusion transformers. CoRR, abs/2406.01125, 2024. URL https://doi.org/10.48550/arXiv.2406.01125

    arXiv doi:10.48550/arxiv.2406.01125 2024

  7. [7]

    Flashattention-2: Faster attention with better parallelism and work partitioning

    Dao, T. Flashattention-2: Faster attention with better parallelism and work partitioning. In The Twelfth International Conference on Learning Representations, 2024. URL https://openreview.net/forum?id=mZn2Xyh9Ec

    2024

  8. [8]

    Flex attention: A programming model for generating optimized attention kernels, 2024

    Dong, J., Feng, B., Guessous, D., Liang, Y., and He, H. Flex attention: A programming model for generating optimized attention kernels, 2024. URL https://arxiv.org/abs/2412.05496

    arXiv 2024

Show all 46 references
  1. [9]

    Scaling rectified flow transformers for high-resolution image synthesis

    Esser, P., Kulal, S., Blattmann, A., Entezari, R., M \"u ller, J., Saini, H., Levi, Y., Lorenz, D., Sauer, A., Boesel, F., et al. Scaling rectified flow transformers for high-resolution image synthesis. In Forty-first International Conference on Machine Learning, 2024

    2024

  2. [10]

    On the content bias in frechet video distance

    Ge, S., Mahapatra, A., Parmar, G., Zhu, J.-Y., and Huang, J.-B. On the content bias in frechet video distance. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp.\ 7277--7288, June 2024

    2024

  3. [11]

    Genmo-Team. Mochi 1. https://github.com/genmoai/models, 2024

    2024

  4. [12]

    Neighborhood attention transformer

    Hassani, A., Walton, S., Li, J., Li, S., and Shi, H. Neighborhood attention transformer. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023

    2023

  5. [13]

    Vbench: Comprehensive benchmark suite for video generative models

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

    2024

  6. [14]

    Hunyuanvideo: A systematic framework for large video generative models, 2025

    HunyuanVideo-Team. Hunyuanvideo: A systematic framework for large video generative models, 2025. URL https://arxiv.org/abs/2412.03603

    2025 arXiv

  7. [15]

    Q., Sablayrolles, A., Mensch, A., Bamford, C., Chaplot, D

    Jiang, A. Q., Sablayrolles, A., Mensch, A., Bamford, C., Chaplot, D. S., de las Casas, D., Bressand, F., Lengyel, G., Lample, G., Saulnier, L., Lavaud, L. R., Lachaux, M.-A., Stock, P., Scao, T. L., Lavril, T., Wang, T., Lacroix, T., and Sayed, W. E. Mistral 7b, 2023. URL http...

    2023 arXiv

  8. [16]

    Gotta go fast when generating data with score-based models

    Jolicoeur-Martineau, A., Li, K., Pich \'e -Taillefer, R., Kachman, T., and Mitliagkas, I. Gotta go fast when generating data with score-based models. arXiv preprint arXiv:2105.14080, 2021

    2021 arXiv

  9. [17]

    Consistency trajectory models: Learning probability flow ode trajectory of diffusion

    Kim, D., Lai, C.-H., Liao, W.-H., Murata, N., Takida, Y., Uesaka, T., He, Y., Mitsufuji, Y., and Ermon, S. Consistency trajectory models: Learning probability flow ode trajectory of diffusion. arXiv preprint arXiv:2310.02279, 2023

    arXiv

  10. [18]

    Li, J., Feng, W., Fu, T.-J., Wang, X., Basu, S., Chen, W., and Wang, W. Y. T2v-turbo: Breaking the quality bottleneck of video consistency model with mixed reward feedback. arXiv preprint arXiv:2405.18750, 2024

    2024 arXiv

  11. [19]

    Open-sora plan: Open-source large video generation model

    Lin, B., Ge, Y., Cheng, X., Li, Z., Zhu, B., Wang, S., He, X., Ye, Y., Yuan, S., Chen, L., et al. Open-sora plan: Open-source large video generation model. arXiv preprint arXiv:2412.00131, 2024

    2024 arXiv

  12. [20]

    L., and Dollár, P

    Lin, T.-Y., Maire, M., Belongie, S., Bourdev, L., Girshick, R., Hays, J., Perona, P., Ramanan, D., Zitnick, C. L., and Dollár, P. Microsoft coco: Common objects in context, 2015. URL https://arxiv.org/abs/1405.0312

    2015 arXiv

  13. [21]

    T., Ben-Hamu, H., Nickel, M., and Le, M

    Lipman, Y., Chen, R. T., Ben-Hamu, H., Nickel, M., and Le, M. Flow matching for generative modeling. arXiv preprint arXiv:2210.02747, 2022

    2022 arXiv

  14. [22]

    Clear: Conv-like linearization revs pre-trained diffusion transformers up, 2024

    Liu, S., Tan, Z., and Wang, X. Clear: Conv-like linearization revs pre-trained diffusion transformers up, 2024. URL https://arxiv.org/abs/2412.16112

    2024 arXiv

  15. [24]

    Swin transformer: Hierarchical vision transformer using shifted windows, 2021 b

    Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., and Guo, B. Swin transformer: Hierarchical vision transformer using shifted windows, 2021 b . URL https://arxiv.org/abs/2103.14030

    2021 arXiv

  16. [25]

    Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps

    Lu, C., Zhou, Y., Bao, F., Chen, J., Li, C., and Zhu, J. Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps. Advances in Neural Information Processing Systems, 35: 0 5775--5787, 2022 a

    2022

  17. [26]

    Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models

    Lu, C., Zhou, Y., Bao, F., Chen, J., Li, C., and Zhu, J. Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models. arXiv preprint arXiv:2211.01095, 2022 b

    2022 arXiv

  18. [27]

    Latte: Latent diffusion transformer for video generation

    Ma, X., Wang, Y., Jia, G., Chen, X., Liu, Z., Li, Y.-F., Chen, C., and Qiao, Y. Latte: Latent diffusion transformer for video generation. arXiv preprint arXiv:2401.03048, 2024

    2024 arXiv

  19. [28]

    Sdedit: Guided image synthesis and editing with stochastic differential equations, 2022

    Meng, C., He, Y., Song, Y., Song, J., Wu, J., Zhu, J.-Y., and Ermon, S. Sdedit: Guided image synthesis and editing with stochastic differential equations, 2022. URL https://arxiv.org/abs/2108.01073

    2022 arXiv

  20. [29]

    Sora, 2024

    OpenAI. Sora, 2024. URL https://openai.com/index/sora/. Accessed: [2024]

    2024

  21. [30]

    and Xie, S

    Peebles, W. and Xie, S. Scalable diffusion models with transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp.\ 4195--4205, 2023

    2023

  22. [31]

    Movie gen: A cast of media foundation models

    Polyak, A., Zohar, A., Brown, A., Tjandra, A., Sinha, A., Lee, A., Vyas, A., Shi, B., Ma, C.-Y., Chuang, C.-Y., et al. Movie gen: A cast of media foundation models. arXiv preprint arXiv:2410.13720, 2024

    2024 arXiv

  23. [32]

    Multistep distillation of diffusion models via moment matching

    Salimans, T., Mensink, T., Heek, J., and Hoogeboom, E. Multistep distillation of diffusion models via moment matching. arXiv preprint arXiv:2406.04103, 2024

    2024 arXiv

  24. [33]

    Adversarial diffusion distillation

    Sauer, A., Lorenz, D., Blattmann, A., and Rombach, R. Adversarial diffusion distillation. arXiv preprint arXiv:2311.17042, 2023

    2023 arXiv

  25. [34]

    Flashattention-3: Fast and accurate attention with asynchrony and low-precision, 2024

    Shah, J., Bikshandi, G., Zhang, Y., Thakkar, V., Ramani, P., and Dao, T. Flashattention-3: Fast and accurate attention with asynchrony and low-precision, 2024. URL https://arxiv.org/abs/2407.08608

    2024 arXiv

  26. [35]

    Make-a-video: Text-to-video generation without text-video data, 2022

    Singer, U., Polyak, A., Hayes, T., Yin, X., An, J., Zhang, S., Hu, Q., Yang, H., Ashual, O., Gafni, O., Parikh, D., Gupta, S., and Taigman, Y. Make-a-video: Text-to-video generation without text-video data, 2022. URL https://arxiv.org/abs/2209.14792

    2022 arXiv

  27. [36]

    Denoising diffusion implicit models

    Song, J., Meng, C., and Ermon, S. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502, 2020

    2010 arXiv

  28. [37]

    and Ermon, S

    Song, Y. and Ermon, S. Generative modeling by estimating gradients of the data distribution. arXiv preprint arXiv:1907.05600, 2019

    1907 arXiv

  29. [38]

    Consistency models

    Song, Y., Dhariwal, P., Chen, M., and Sutskever, I. Consistency models. arXiv preprint arXiv:2303.01469, 2023

    2023 arXiv

  30. [39]

    F., Arora, S., Singhal, A., Fu, D

    Spector, B. F., Arora, S., Singhal, A., Fu, D. Y., and Ré, C. Thunderkittens: Simple, fast, and adorable ai kernels, 2024. URL https://arxiv.org/abs/2410.20399

    2024 arXiv

  31. [40]

    Z., Khabsa, M., Fang, H., and Ma, H

    Wang, S., Li, B. Z., Khabsa, M., Fang, H., and Ma, H. Linformer: Self-attention with linear complexity. arXiv preprint arXiv:2006.04768, 2020

    2006 arXiv

  32. [41]

    Lavie: High-quality video generation with cascaded latent diffusion models

    Wang, Y., Chen, X., Ma, X., Zhou, S., Huang, Z., Wang, Y., Yang, C., He, Y., Yu, J., Yang, P., et al. Lavie: High-quality video generation with cascaded latent diffusion models. arXiv preprint arXiv:2309.15103, 2023

    2023 arXiv

  33. [42]

    Mlcm: Multistep consistency distillation of latent diffusion model

    Xie, Q., Liao, Z., Deng, Z., Tang, S., Lu, H., et al. Mlcm: Multistep consistency distillation of latent diffusion model. arXiv preprint arXiv:2406.05768, 2024

    2024 arXiv

  34. [43]

    Gated linear attention transformers with hardware-efficient training

    Yang, S., Wang, B., Shen, Y., Panda, R., and Kim, Y. Gated linear attention transformers with hardware-efficient training. arXiv preprint arXiv:2401.00002, 2024 a

    2024 arXiv

  35. [44]

    Cogvideox: Text-to-video diffusion models with an expert transformer

    Yang, Z., Teng, J., Zheng, W., Ding, M., Huang, S., Xu, J., Yang, Y., Hong, W., Zhang, X., Feng, G., et al. Cogvideox: Text-to-video diffusion models with an expert transformer. arXiv preprint arXiv:2408.06072, 2024 b

    2024 arXiv

  36. [45]

    T., and Park, T

    Yin, T., Gharbi, M., Zhang, R., Shechtman, E., Durand, F., Freeman, W. T., and Park, T. One-step diffusion with distribution matching distillation. arXiv preprint arXiv:2311.18828, 2023

    2023 arXiv

  37. [46]

    T., and Park, T

    Yin, T., Gharbi, M., Zhang, R., Shechtman, E., Durand, F., Freeman, W. T., and Park, T. One-step diffusion with distribution matching distillation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 6613--6623, 2024

    2024

  38. [47]

    Open-sora: Democratizing efficient video production for all, March 2024

    Zheng, Z., Peng, X., Yang, T., Shen, C., Li, S., Liu, H., Zhou, Y., Li, T., and You, Y. Open-sora: Democratizing efficient video production for all, March 2024. URL https://github.com/hpcaitech/Open-Sora

    2024

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.