Pith. sign in

REVIEW 4 major objections 5 minor 17 cited by

PixNerd: Pixel Neural Field Diffusion

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read PixNerd claims that a pixel diffusion transformer with a per-patch neural field decoder reaches 2.15 FID on ImageNet 256x256 and 2.84 FID on ImageNet 512x512 without a VAE or cascade pipeline.

desk verdict PixNerd's neural-field decoder is genuinely new and the ablations support it, but the manuscript's internal inconsistencies (an 'abc steps' placeholder and DPG score reported as both 0.82 and 80.9) make the headline numbers unverifiable as written. read the letter →

arxiv 2507.23268 v2 pith:GI57ZYXZ submitted 2025-07-31 cs.CV

Shuai Wang , Ziteng Gao , Chenhui Zhu , Weilin Huang , Limin Wang This is my paper
classification cs.CV
keywords diffusiontransformerpixel-spaceneuralfieldimplicitrepresentationimagegenerationclass-conditionaltext-to-imageVAE-free
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

Diffusion transformers currently owe most of their image quality to a pre-trained variational autoencoder (VAE) that compresses pixels into a latent space; this paper argues that the VAE is a workaround, not a requirement. PixNerd is a single-scale, single-stage diffusion transformer that denoises raw pixels directly, using patch size 16 to keep the token count the same as a latent model. The authors report 2.15 FID on ImageNet 256x256 and 2.84 FID on ImageNet 512x512 with no VAE, no cascade pipeline, and no multi-scale training, bringing the quality close to latent DiT-class models under similar compute. The claim matters because it shifts the difficulty of pixel diffusion from "the space is too hard" to "the decoder was too weak": the paper's evidence is that replacing the final linear projection with a per-patch neural field lowers training loss and closes most of the FID gap. The same design is extended to text-to-image generation, where it reports 0.73 on GenEval and 80.9 on DPG.

What carries the argument

The load-bearing mechanism is the patch-wise neural field decoder. Given the last hidden state $X_n$ of patch $n$, two linear layers predict the row-normalized weights $\{W_1^n, W_2^n\}$ of a two-layer MLP (Eq. 9). The MLP is evaluated at every pixel $(i,j)$ inside the patch: its input is the concatenation of a DCT coordinate encoding $\cos(k_1 i)\cos(k_2 j)$ for $k_1,k_2\in(0,K]$ and the noisy pixel value $x_n(i,j)$, and its output is passed through a final linear layer to give the velocity $v_n(i,j)$ (Eqs. 10-11). Because the decoder is a coordinate-based function, it can be queried at arbitrary locations, which is what enables training-free arbitrary-resolution generation: the token grid stays fixed at the pretraining resolution while the neural field coordinates are interpolated.

What would settle it

Take the trained PixNerd-L/16 model and evaluate it with the DCT coordinate encoding replaced by zeros, so the decoder sees only the noisy pixel value; if FID at 400k training steps stays close to the full model, then coordinates are doing little and the claimed mechanism is not the source of the gain. Conversely, scale the neural field MLP to four layers and 128 channels: if the FID curve does not improve beyond the 2-layer/64-channel default, then the simple-MLP capacity assumption is not the binding constraint.

Watch

Extended reading notes

Core claim

The central claim is that a large-patch pixel-space diffusion transformer can be competitive with latent-space diffusion if the patch decoder is a coordinate-conditioned neural field rather than a linear layer. Concretely, the transformer's last hidden state for a 16x16 patch predicts the weights of a two-layer MLP; that MLP then takes, for each pixel in the patch, a DCT-basis encoding of the pixel's local coordinates concatenated with the noisy pixel value, and outputs the diffusion velocity. This gives the decoder an explicit sense of spatial locality that a linear projection on flattened patches does not have, which the paper identifies as the reason it can decode large patches without a cascade. The paper presents this as evidence that pixel-space diffusion is viable as a single-scale, single-stage, end-to-end system, and that the VAE's two-stage training is an optional shortcut rather than a structural necessity.

Load-bearing premise

The load-bearing premise is that a two-layer MLP whose weights are predicted per patch can faithfully decode the velocity of all 256 pixels inside a 16x16 patch from just a coordinate encoding and the noisy pixel value, and the paper shows ablations over MLP width, depth, and normalization but does not prove that this decoder class is expressive enough to be the right one.

Editorial extensions

If this is right

  • A single-stage pixel-space diffusion model is computationally practical at ImageNet scale: with patch size 16 it keeps the same token count as a latent DiT, and the paper reports roughly 8x lower per-step latency than ADM-G and PixelFlow while training faster per iteration than a latent SiT-L/2 baseline.
  • The model reaches its headline image quality without any cascade: 2.15 FID on ImageNet 256x256 with 100 Euler steps, and 2.84 FID on ImageNet 512x512 after fine-tuning from the 256x256 checkpoint.
  • Large patches stop being a barrier to pixel-space learning: at identical training steps, PixNerd-L/16 reports consistently lower flow-matching loss and lower DINOv2 representation-alignment loss than the linear-projection Baseline-L/16.
  • Arbitrary-resolution generation is a by-product of the coordinate decoder: keeping the number of tokens constant and interpolating neural field coordinates produces multi-resolution outputs without fine-tuning.
  • The same decoder transfers to text-to-image: PixNerd-XXL/16 reaches 0.73 on GenEval and 80.9 on DPG with a 45M-image training set, outperforming the previous pixel-space PixelFlow on both benchmarks in the paper's tables.

Reading between the lines

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

  • An implication the paper leaves implicit is that the same coordinate-based decoder should transfer to other dense-output domains, such as video or 3D, where a transformer emits patch tokens that must be expanded into many output points.
  • A testable extension the paper does not run is to measure resolution extrapolation with spectral statistics; if the coordinate interpolation preserves low frequencies but loses high-frequency structure, the claimed training-free arbitrary resolution has a boundary.
  • The paper's comparative framing implies a stronger claim than the one it proves: that the VAE contributes artifacts rather than being a necessary compression. A direct test would be to fine-tune PixNerd at 512x512 with a high-resolution post-training stage and compare against a latent model post-trained identically.
  • The paper itself concedes some generated samples show unclear details, so the fair reading is that the gap to latent models has narrowed enough to make the VAE optional, not that pixel space has become strictly better.
Share X Bluesky LinkedIn Reddit HN

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. PixNerd proposes a pixel-space diffusion transformer in which the final linear patch decoder is replaced by a per-patch neural field. The transformer's last hidden state for each patch is linearly projected into MLP weights (Eq. 9), and each pixel's diffusion velocity is decoded from DCT coordinate encodings plus the noisy pixel value (Eq. 10). The model is trained end-to-end at a single scale with patch size 16, avoiding VAE compression and cascade pipelines. On class-conditional ImageNet, PixNerd-XL/16 reports FID 2.15 at 256×256 and 2.84 at 512×512; on text-to-image, PixNerd-XXL/16 reports GenEval 0.73 and DPG 80.9. The paper includes a controlled comparison against a linear-decoder baseline (Baseline-L/16) and component ablations supporting the neural-field design.

Significance. If the results hold, PixNerd is a significant advance for pixel-space generation: it shows that a single-scale DiT with patch size 16 can approach latent DiT performance without a VAE, and the resource comparison in Table 1 suggests large latency gains over cascade pixel models. The controlled Baseline-L/16 comparison and the loss curves in Fig. 6 provide direct evidence that the neural-field decoder is responsible for the improvement, and the ablations in Fig. 7 give useful design guidance. The method is simple and does not rely on adversarial VAE training. However, the empirical claims are currently undermined by incomplete training details and inconsistent benchmark numbers, so the result cannot yet be fully assessed.

major comments (4)
  1. [§4.4, Table 3] The 512×512 result is not reproducible as reported. The text in §4.4 says the ImageNet 512 model is fine-tuned from the 256 model 'for abc steps' — a literal placeholder — and states the guidance interval is [0.3, 1.0], while the caption of Table 3 says interval [0.1, 1] with CFG 3.5. Please provide the actual step/epoch count, reconcile the interval, and state whether the reported FID was selected on a validation set or is a single test-set evaluation. The manuscript lists repository URLs but does not state whether code or checkpoints will be available, so these ambiguities cannot be resolved from the paper.
  2. [§4.5, DPG paragraph] The DPG result is internally inconsistent: the text says PixNerd-XXL/16 'achieves 0.82 overall score', while the abstract and Table 5 report 80.9. Since all DPG scores in Table 5 are on a 0–100 scale, 80.9 is presumably the correct value, but the manuscript currently states two different headline numbers for the same benchmark. Correct all occurrences and double-check the GenEval/DPG tables for further such discrepancies.
  3. [§4.3, Fig. 7e and Table 2] The headline FID numbers are obtained after tuning hyperparameters on the target metric: Fig. 7e sweeps CFG values 3.0–5.0 and picks 3.4–3.6 on FID10k, and the paper then chooses 3.5 as the default for the FID50k results in Tables 2 and 3. This is selection on the evaluation benchmark and makes the reported 2.15/2.84 FID optimistic. Please report results for a fixed CFG/interval (or a validation split), and clearly state that the final numbers use the tuned schedule.
  4. [Eqs. (9)–(11)] The central architectural assumption is that a linear projection of the patch hidden state predicts neural-field weights sufficient to decode all pixels in a 16×16 patch, with DCT coordinates plus the noisy pixel value as complete local conditioning. The Baseline-L/16 comparison and the width/depth ablations support this choice empirically, but they do not test the capacity ceiling of the decoder. A concrete upper-bound experiment (e.g., larger MLP hidden dimension, deeper MLP, or a small transformer-based per-patch decoder) would clarify how much of the remaining gap to latent DiT is due to decoder capacity.
minor comments (5)
  1. [§1, Fig. 5] The paper contains several typos, including 'Pixel Nerual FieldDiffusion' in the introduction, 'Basline-L/16' in Fig. 5, and a stray space in 'V AE' in multiple places; please proofread.
  2. [§4.3, §4.5, Fig. 3] The solver is referred to as 'Adam2', 'Adams-2nd', and 'Adams-2' interchangeably in different parts of the paper; please use one consistent name.
  3. [§3.1, Eqs. (2), (7), (12)] The definitions of f(t) and g(t) are garbled: the expression for g(t) lacks the standard square-root/brace structure, and Eq. (7) prints '20πi' where superscripts such as 2^0πi appear intended. Eq. (12) should also define the range and normalization of k1 and k2.
  4. [Fig. 7e] The text says CFG was swept from 3.0 to 5.0, but the horizontal axis only shows 3.2–4.8; please clarify the exact range and state whether the FID10k values are single runs or averaged over multiple seeds.
  5. [Table 1] The statement that PixNerd is 'nearly 8× faster than other pixel diffusion models' is not directly supported by the table: ADM-G is 4.21s versus 0.65s (about 6.5×) and PixelFlow is 10.1s versus 0.65s (about 15.5×). Please specify the hardware and clarify which comparison the factor refers to.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: PixelNerd's neural-field decoder is compared against external benchmarks and its equations do not presuppose the headline results.

full rationale

The paper's load-bearing claim is empirical: replacing the final linear projection of a pixel-space diffusion transformer with a per-patch neural-field decoder (Eqs. 9-11) yields competitive FID on ImageNet and strong scores on GenEval and DPG. This claim is checked against external benchmarks and against an internal Baseline-L/16 trained with identical optimizer settings. The velocity target v_t = x_real - epsilon is defined independently of the network, and the decoder is trained with a standard flow-matching loss; the predicted MLP weights in Eq. 9 are learned end-to-end from the training objective, not fitted to the reported metrics. The CFG, interval, solver, and neural-field ablations in Figure 7 are hyperparameter/model selection rather than circular derivation. The self-citations [15,18] appear as architecture and method pointers alongside external works such as DiT, SiT, REPA, and interval guidance; they are not the load-bearing justification for the neural-field decoding scheme. The concerns flagged in the manuscript text—the placeholder 'abc steps' for the ImageNet 512 fine-tuning in Section 4.4 and the inconsistent DPG score reported as both 0.82 and 80.9—are reproducibility and correctness issues, not circularity, because they do not make the derivation equivalent to its inputs. No circular steps are therefore identified.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on several empirically selected hyperparameters (CFG, interval guidance, MLP width and depth) and on the untested-in-theory assumption that a predicted per-patch MLP can decode velocity from coordinates and a single noisy pixel value. These are the main costs the reader must accept.

free parameters (5)
  • Classifier-free guidance (CFG) value = 3.5
    Selected by sweeping CFG from 3.0 to 5.0 and taking the best FID10k on ImageNet 256 (Section 4.3, Fig. 7e).
  • Interval guidance bounds = [0.1, 1.0] for 256, [0.3, 1.0] for 512
    Tuned per resolution; 512 uses a different interval to reach 2.84 FID (Section 4.4, Table 3).
  • Neural field hidden channels = 64
    Chosen from {36, 64, 72} based on FID50K and compute trade-off (Section 4.2, Fig. 7b).
  • Neural field MLP depth = 2
    Chosen from {1, 2, 4} based on FID50K and compute trade-off (Section 4.2, Fig. 7c).
  • Representation alignment loss weight = 0.5
    Set to 0.5 to align 8th-layer features with DINOv2-Base (Section 4, Training Details).
assumptions (5)
  • standard math Rectified flow velocity objective equals the difference between clean data and noise, and the reverse-time diffusion and ODE theory holds (Eqs. 1-4).
    Basis for training and sampling; standard result from Anderson (1982) and the flow matching literature.
  • domain assumption Transformer blocks with AdaLN, SwiGLU, RoPE, and RMSNorm are effective and stable for diffusion training.
    Borrowed from prior work; the paper does not re-derive this.
  • domain assumption Representation alignment to DINOv2 features improves diffusion transformer training.
    Adopted from REPA; the paper uses a loss weight of 0.5 without justification beyond prior work.
  • ad hoc to paper A linear projection from the patch hidden state is sufficient to predict neural field weights that decode pixel velocities (Eq. 9).
    Core design assumption of the paper; validated only empirically by ablations, not by derivation.
  • ad hoc to paper DCT coordinate encoding plus the noisy pixel value is a complete local conditioning set for velocity decoding (Eq. 10).
    The paper abandons sine and cosine encoding in favor of DCT after empirical comparison, but provides no proof that this encoding is sufficient.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PixNerd: Pixel Neural Field Diffusion." pith.science (2026). https://pith.science/paper/GI57ZYXZ

@misc{pith2026250723268,
  author       = {Pith},
  title        = {Pith review of: PixNerd: Pixel Neural Field Diffusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GI57ZYXZ}},
  note         = {Machine review of arXiv:2507.23268}
}
abstract

The current success of diffusion transformers heavily depends on the compressed latent space shaped by the pre-trained variational autoencoder(VAE). However, this two-stage training paradigm inevitably introduces accumulated errors and decoding artifacts. To address the aforementioned problems, researchers return to pixel space at the cost of complicated cascade pipelines and increased token complexity. In contrast to their efforts, we propose to model the patch-wise decoding with neural field and present a single-scale, single-stage, efficient, end-to-end solution, coined as pixel neural field diffusion~(PixelNerd). Thanks to the efficient neural field representation in PixNerd, we directly achieved 2.15 FID on ImageNet $256\times256$ and 2.84 FID on ImageNet $512\times512$ without any complex cascade pipeline or VAE. We also extend our PixNerd framework to text-to-image applications. Our PixNerd-XXL/16 achieved a competitive 0.73 overall score on the GenEval benchmark and 80.9 overall score on the DPG benchmark.

Figures

Figures reproduced from arXiv: 2507.23268 by the authors.

Figure 1
Figure 1. Left: Comparison with other diffusion models. Our LargePatch/SingleScale pixel space diffusion keeps consistent tokens as latent diffusion among diffusion steps. Right: PixNerd architecture. PixNerd follows the diffusion transformer design, replacing the final linear projection with a neural field to model the large patch details. Abstract The current success of diffusion transformers heavily depends on the compress… view at source ↗
Figure 2
Figure 2. Selected 256 × 256 and 512 × 512 resolution samples. Generated from PixNerd-XL/16 trained on ImageNet 256 × 256 resolution and ImageNet 512 × 512 resolution with CFG = 3.5. Surreal landscape where a waterfall flows upward into a floating island, lush vegetation clinging to the cliffs, and a rainbow arching overhead. Bustling Moroccan marketplace at sunset: colorful rugs hang from stone archways, vendors sell spices … view at source ↗
Figure 3
Figure 3. The Text-to-Image 512 × 512 visualization with text descriptions of different lengths and styles. Given text descriptions of different lengths and styles, PixNerd can generate promising samples with a large patch size of 16. We used Adams-2nd solver with 25 steps and a CFG value of 4.0 for sampling. 2 [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Training-free arbitrary resolution generation. We keep the amount of tokens in PixNerd as constant as pretraining resolution, we only interpolate the neural field coordinates for different resolutions to yield multi-resolution images. VAEs typically requires adversaria…
Figure 5
Figure 5. Figure 5: The visualization Comparison with Baseline-L/16 under 400k training steps.With the help of neural field representation, our PixNerd-L/16 yields promising details and better structure. 4 Experiments We conduct ablation studies and baseline comparison experiments on Imag…
Figure 6
Figure 6. Figure 6: Loss Comparison with Diffusion Transformer Baselines. Our PixNerd-L/16 achieves consistently lower REPA loss and flow matching loss than its diffusion transformer counterpart. 100 200 300 400 Training iterations(K steps) 30 35 40 45 50 55 FID50K Norm Fc1/Fc2 Norm Fc1/F…
Figure 7
Figure 7. Figure 7: Ablations studies of PixNerd. We conduct ablation studies on class-to-image generation benchmark ImageNet256 × 256 with PixNerd-L/16. 4.2 Neural Field Design We conduct ablation studies on PixNerd-L/16, which comprises 22 transformer layers with 1024 channels. The Neur…
Figure 8
Figure 8. Figure 8: The Text-to-Image 512 × 512 visualization with different solvers. We armed PixNerd with different ODE solvers, eg, Euler, Adams-2nd, Adams-3rd. Adams solver achieves better visual quality than the naive Euler solver. Also, thanks to the powerful text embedding in Qwen3…

Discussion (0). Sign in to comment.

Forward citations

Cited by 17 Pith papers

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

  1. A Frozen Pixel-Space Diffusion Model Can Guide Itself with Its Own Samples

    cs.CV 2026-07 conditional novelty 7.0 of 10

    Training a small adapter on a frozen pixel diffusion model's own samples and extrapolating the final prediction away from the adapter's intermediate prediction improves FID on ImageNet.

  2. SPEED: One-Step Pixel Diffusion for High-quality Video Frame Interpolation

    cs.MM 2026-07 conditional novelty 7.0 of 10

    SPEED generates an interpolated video frame in a single pixel-space diffusion step, reporting state-of-the-art LPIPS on DAVIS, SNU-FILM, and XTest4K with lower latency and memory than latent-diffusion baselines.

  3. Representation Forcing for Bottleneck-Free Unified Multimodal Models

    cs.CV 2026-05 unverdicted novelty 6.5 of 10

    Representation Forcing lets a UMM decoder autoregressively predict its own understanding representations as in-context tokens that guide pixel-space diffusion, matching VAE-based generation without an external latent space.

  4. Beyond Token-Level Cross-Entropy: Fr\'echet Distributional Post-Training for Autoregressive Image Generation

    cs.CV 2026-08 conditional novelty 6.0 of 10

    FD-loss post-training with detached rollout replay and a probability-level straight-through estimator improves FID and FD_r6 across eight ImageNet configurations.

  5. Stable Autoregressive Speech Generation with Low-Frame-Rate High-Dimensional Continuous Tokens

    eess.AS 2026-07 conditional novelty 6.0 of 10

    Autoregressive TTS from 8-Hz, 768-dimensional continuous tokens works when the tokenizer shapes its latent space with a low-dimensional core and an energy hierarchy, and the generator separates guidance into local, se...

  6. WaiT for the Signal: Simple Frequency-Aware Flow-Matching

    cs.CV 2026-07 conditional novelty 6.0 of 10

    WaiT delays high-frequency wavelet bands in flow-matching image generation until coarse structure emerges, improving quality and cutting compute, with a reported SOTA FID of 1.30 on ImageNet 512.

  7. Amortized Moment Matching for Visual Generation

    cs.LG 2026-07 accept novelty 6.0 of 10

    Amortized Fréchet Distance uses neural nets to match conditional means and covariances, yielding stronger one-step visual generators than explicit FD-loss or multi-step teachers.

  8. Twins: Learn to Predict Unified Representations with Focal Loss

    cs.CV 2026-07 conditional novelty 6.0 of 10

    Channel-wise concatenation of SigLIP2 and Flux VAE features into one token, trained with a focal-style flow-matching loss, yields a unified representation with 1.59 gFID on ImageNet 256 and VAE-level reconstruction.

  9. DuSPiT: Dual-Branch Sub-Patch Pixel Diffusion Transformer

    cs.CV 2026-07 conditional novelty 6.0 of 10

    DuSPiT splits pixel diffusion into a compact global-structure branch and a high-capacity subpatch detail branch, reaching ImageNet-512 FID 1.52 at 329 GFLOPs, below JiT-G/32 (1.78).

  10. PixGS: Pixel-Space Diffusion for Direct 3D Gaussian Splat Generation

    cs.CV 2026-07 unverdicted novelty 6.0 of 10

    PixGS is a single-stage pixel-space diffusion model that directly produces high-quality 3D Gaussian Splats from text or images in ~1s, outperforming multi-stage latent methods on standard benchmarks.

  11. SRC-Flow: Compact Semantic Representations Enable Normalizing Flows for Image Generation

    cs.CV 2026-05 unverdicted novelty 6.0 of 10

    SRC-Flow compresses RAE features via a Semantic Representation Compressor into a low-dimensional space, enabling normalizing flows to reach gFID 1.65 on ImageNet 256x256 and 2.07 on 512x512 while retaining exact likelihoods.

  12. FREPix: Frequency-Heterogeneous Flow Matching for Pixel-Space Image Generation

    cs.CV 2026-05 unverdicted novelty 6.0 of 10

    FREPix generates images by transporting low- and high-frequency wavelet components along separate schedules, reaching 1.91 FID on ImageNet 256×256.

  13. MixFlow Training: Alleviating Exposure Bias with Slowed Interpolation Mixture

    cs.CV 2025-12 conditional novelty 6.0 of 10

    Training diffusion models on a mixture of higher-noise interpolations (MixFlow) improves generation FID across SiT, REPA, RAE and SD3.5, reaching ImageNet 256 gFID 1.43 after post-training.

  14. STEREOFLOW: Progressive Stereo Matching with StereoDiT and Transition Flow Matching

    cs.CV 2026-07 conditional novelty 5.0 of 10

    A hybrid stereo-matching model uses a cascade matching network to propose disparities and a diffusion transformer to refine ambiguous regions; it claims state-of-the-art benchmark results.

  15. Learning on the Manifold: Unlocking Standard Diffusion Transformers with Representation Encoders

    cs.LG 2026-02 conditional novelty 5.0 of 10

    Training flow matching along sphere geodesics with a curvature-aware loss weight lets standard DiT-B converge on DINOv2 features (FID 3.37 with guidance), contradicting the need for width scaling.

  16. Rethinking Pixel Mean Flows via Interval Denoiser

    cs.CV 2026-08 conditional novelty 4.0 of 10

    A new Interval Denoiser target gives 4.55 FID at 1 step and 3.98 FID at 2 steps for pixel-space ImageNet 256 generation, without perceptual losses.

  17. Pixel-Space Diffusion Transformers

    cs.CV 2026-07 conditional novelty 3.0 of 10

    A systematic review of pixel-space diffusion transformers, categorizing architectures and challenges for end-to-end image generation without latent compression.

Reference graph

Works this paper leans on

78 extracted references · 35 canonical work pages · cited by 17 Pith papers

  1. [1]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 1, 3, 4, 5, 8, 10

    work page 2022

  2. [2]

    Reconstruction vs

    Jingfeng Yao and Xinggang Wang. Reconstruction vs. generation: Taming optimization dilemma in latent diffusion models.arXiv preprint arXiv:2501.01423, 2025. 1, 3, 4, 5, 6, 8

    arXiv 2025

  3. [3]

    Deep compression autoencoder for efficient high-resolution diffusion models.arXiv preprint arXiv:2410.10733, 2024

    Junyu Chen, Han Cai, Junsong Chen, Enze Xie, Shang Yang, Haotian Tang, Muyang Li, Yao Lu, and Song Han. Deep compression autoencoder for efficient high-resolution diffusion models.arXiv preprint arXiv:2410.10733, 2024. 1, 3, 4, 5

    arXiv 2024

  4. [4]

    Large scale gan training for high fidelity natural image synthesis.arXiv preprint arXiv:1809.11096, 2018

    Andrew Brock, Jeff Donahue, and Karen Simonyan. Large scale gan training for high fidelity natural image synthesis.arXiv preprint arXiv:1809.11096, 2018. 3

    arXiv 2018

  5. [5]

    Stylegan-xl: Scaling stylegan to large diverse datasets

    Axel Sauer, Katja Schwarz, and Andreas Geiger. Stylegan-xl: Scaling stylegan to large diverse datasets. In ACM SIGGRAPH 2022 conference proceedings, pages 1–10, 2022. 3

    work page 2022

  6. [6]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shechtman, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 586–595, 2018. 3

    work page 2018

  7. [7]

    Repa-e: Unlocking vae for end-to-end tuning with latent diffusion transformers.arXiv preprint arXiv:2504.10483,

    Xingjian Leng, Jaskirat Singh, Yunzhong Hou, Zhenchang Xing, Saining Xie, and Liang Zheng. Repa-e: Unlocking vae for end-to-end tuning with latent diffusion transformers.arXiv preprint arXiv:2504.10483,

    arXiv

  8. [8]

    Diffusion models beat gans on image synthesis.Advances in neural information processing systems, 34:8780–8794, 2021

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis.Advances in neural information processing systems, 34:8780–8794, 2021. 3, 4, 5, 6, 8, 9

    work page 2021

Show all 78 references
  1. [9]

    Pixelflow: Pixel-space generative models with flow.arXiv preprint arXiv:2504.07963, 2025

    Shoufa Chen, Chongjian Ge, Shilong Zhang, Peize Sun, and Ping Luo. Pixelflow: Pixel-space generative models with flow.arXiv preprint arXiv:2504.07963, 2025. 3, 4, 6, 8, 10, 11

    2025 arXiv

  2. [10]

    Relay diffusion: Unifying diffusion process across resolutions for image synthesis.arXiv preprint arXiv:2309.03350, 2023

    Jiayan Teng, Wendi Zheng, Ming Ding, Wenyi Hong, Jianqiao Wangni, Zhuoyi Yang, and Jie Tang. Relay diffusion: Unifying diffusion process across resolutions for image synthesis.arXiv preprint arXiv:2309.03350, 2023. 3, 4, 8

    2023 arXiv

  3. [11]

    Nerf: Representing scenes as neural radiance fields for view synthesis.Communications of the ACM, 65(1):99–106, 2021

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis.Communications of the ACM, 65(1):99–106, 2021. 3, 5

    2021

  4. [12]

    Implicit neural representations with periodic activation functions.Advances in neural information processing systems, 33:7462–7473, 2020

    Vincent Sitzmann, Julien Martel, Alexander Bergman, David Lindell, and Gordon Wetzstein. Implicit neural representations with periodic activation functions.Advances in neural information processing systems, 33:7462–7473, 2020. 3, 5

    2020

  5. [13]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 4195–4205, 2023. 4, 5, 8, 9 12

    2023

  6. [14]

    Sit: Exploring flow and diffusion-based generative models with scalable interpolant transformers.arXiv preprint arXiv:2401.08740, 2024

    Nanye Ma, Mark Goldstein, Michael S Albergo, Nicholas M Boffi, Eric Vanden-Eijnden, and Saining Xie. Sit: Exploring flow and diffusion-based generative models with scalable interpolant transformers.arXiv preprint arXiv:2401.08740, 2024. 4, 5, 8, 9

    2024 arXiv

  7. [15]

    Decoupled diffusion transformer.arXiv preprint arXiv:2504.05741, 2025

    Shuai Wang, Zhi Tian, Weilin Huang, and Limin Wang. Decoupled diffusion transformer.arXiv preprint arXiv:2504.05741, 2025. 4, 5, 6, 8, 9

    2025 arXiv

  8. [16]

    Analyzing and improving the training dynamics of diffusion models

    Tero Karras, Miika Aittala, Jaakko Lehtinen, Janne Hellsten, Timo Aila, and Samuli Laine. Analyzing and improving the training dynamics of diffusion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 24174–24184, 2024. 4

    2024

  9. [17]

    Diffusion models need visual priors for image generation.arXiv preprint arXiv:2410.08531, 2024

    Xiaoyu Yue, Zidong Wang, Zeyu Lu, Shuyang Sun, Meng Wei, Wanli Ouyang, Lei Bai, and Luping Zhou. Diffusion models need visual priors for image generation.arXiv preprint arXiv:2410.08531, 2024. 4

    2024 arXiv

  10. [18]

    Exploring dcn-like architecture for fast image generation with arbitrary resolution.Advances in Neural Information Processing Systems, 37:87959–87977, 2024

    Shuai Wang, Zexian Li, Tianhui Song, Xubin Li, Tiezheng Ge, Bo Zheng, and Limin Wang. Exploring dcn-like architecture for fast image generation with arbitrary resolution.Advances in Neural Information Processing Systems, 37:87959–87977, 2024. 4, 8, 9

    2024

  11. [19]

    Dim: Diffusion mamba for efficient high-resolution image synthesis.arXiv preprint arXiv:2405.14224, 2024

    Yao Teng, Yue Wu, Han Shi, Xuefei Ning, Guohao Dai, Yu Wang, Zhenguo Li, and Xihui Liu. Dim: Diffusion mamba for efficient high-resolution image synthesis.arXiv preprint arXiv:2405.14224, 2024. 4

    2024 arXiv

  12. [20]

    Dmm: Building a versatile image generation model via distillation-based model merging.arXiv preprint arXiv:2504.12364, 2025

    Tianhui Song, Weixin Feng, Shuai Wang, Xubin Li, Tiezheng Ge, Bo Zheng, and Limin Wang. Dmm: Building a versatile image generation model via distillation-based model merging.arXiv preprint arXiv:2504.12364, 2025. 4

    2025 arXiv

  13. [21]

    Score identity distillation: Exponentially fast distillation of pretrained diffusion models for one-step generation

    Mingyuan Zhou, Huangjie Zheng, Zhendong Wang, Mingzhang Yin, and Hai Huang. Score identity distillation: Exponentially fast distillation of pretrained diffusion models for one-step generation. In Forty-first International Conference on Machine Learning, 2024. 4

    2024

  14. [22]

    All are worth words: A vit backbone for diffusion models

    Fan Bao, Shen Nie, Kaiwen Xue, Yue Cao, Chongxuan Li, Hang Su, and Jun Zhu. All are worth words: A vit backbone for diffusion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 22669–22679, 2023. 4, 5

    2023

  15. [23]

    Understanding diffusion objectives as the elbo with simple data augmentation.Advances in Neural Information Processing Systems, 36:65484–65516, 2023

    Diederik Kingma and Ruiqi Gao. Understanding diffusion objectives as the elbo with simple data augmentation.Advances in Neural Information Processing Systems, 36:65484–65516, 2023. 4

    2023

  16. [24]

    simple diffusion: End-to-end diffusion for high resolution images

    Emiel Hoogeboom, Jonathan Heek, and Tim Salimans. simple diffusion: End-to-end diffusion for high resolution images. InInternational Conference on Machine Learning, pages 13213–13232. PMLR, 2023. 4, 9

    2023

  17. [25]

    Fractal generative models.arXiv preprint arXiv:2502.17437, 2025

    Tianhong Li, Qinyi Sun, Lijie Fan, and Kaiming He. Fractal generative models.arXiv preprint arXiv:2502.17437, 2025. 4, 8, 9

    2025 arXiv

  18. [26]

    Normalizing flows are capable generative models.arXiv preprint arXiv:2412.06329, 2024

    Shuangfei Zhai, Ruixiang Zhang, Preetum Nakkiran, David Berthelot, Jiatao Gu, Huangjie Zheng, Tianrong Chen, Miguel Angel Bautista, Navdeep Jaitly, and Josh Susskind. Normalizing flows are capable generative models.arXiv preprint arXiv:2412.06329, 2024. 4

    2024 arXiv

  19. [27]

    Reverse-time diffusion equation models.Stochastic Processes and their Applications, 12(3):313–326, 1982

    Brian DO Anderson. Reverse-time diffusion equation models.Stochastic Processes and their Applications, 12(3):313–326, 1982. 4

    1982

  20. [28]

    Score-based generative modeling through stochastic differential equations.arXiv preprint arXiv:2011.13456, 2020

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations.arXiv preprint arXiv:2011.13456, 2020. 4

    2011 arXiv

  21. [29]

    Elucidating the design space of diffusion-based generative models.Advances in Neural Information Processing Systems, 35:26565–26577, 2022

    Tero Karras, Miika Aittala, Timo Aila, and Samuli Laine. Elucidating the design space of diffusion-based generative models.Advances in Neural Information Processing Systems, 35:26565–26577, 2022. 4, 9

    2022

  22. [30]

    Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020. 4

    2020

  23. [31]

    Llama: Open and efficient foundation language models.arXiv preprint arXiv:2302.13971, 2023

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundation language models.arXiv preprint arXiv:2302.13971, 2023. 5, 6

    2023 arXiv

  24. [32]

    Llama 2: Open foundation and fine-tuned chat models.arXiv preprint arXiv:2307.09288, 2023

    Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. Llama 2: Open foundation and fine-tuned chat models.arXiv preprint arXiv:2307.09288, 2023. 5, 6 13

    2023 arXiv

  25. [33]

    Roformer: Enhanced transformer with rotary position embedding.Neurocomputing, 568:127063, 2024

    Jianlin Su, Murtadha Ahmed, Yu Lu, Shengfeng Pan, Wen Bo, and Yunfeng Liu. Roformer: Enhanced transformer with rotary position embedding.Neurocomputing, 568:127063, 2024. 5, 6

    2024

  26. [34]

    Visionllama: A unified llama interface for vision tasks.arXiv preprint arXiv:2403.00522, 2024

    Xiangxiang Chu, Jianlin Su, Bo Zhang, and Chunhua Shen. Visionllama: A unified llama interface for vision tasks.arXiv preprint arXiv:2403.00522, 2024. 5, 6

    2024 arXiv

  27. [35]

    Fit: Flexible vision transformer for diffusion model.arXiv preprint arXiv:2402.12376, 2024

    Zeyu Lu, Zidong Wang, Di Huang, Chengyue Wu, Xihui Liu, Wanli Ouyang, and Lei Bai. Fit: Flexible vision transformer for diffusion model.arXiv preprint arXiv:2402.12376, 2024. 5, 6

    2024 arXiv

  28. [36]

    Seedream 2.0: A native chinese-english bilingual image generation foundation model

    Lixue Gong, Xiaoxia Hou, Fanshi Li, Liang Li, Xiaochen Lian, Fei Liu, Liyang Liu, Wei Liu, Wei Lu, Yichun Shi, et al. Seedream 2.0: A native chinese-english bilingual image generation foundation model. arXiv preprint arXiv:2503.07703, 2025. 5, 9

    2025 arXiv

  29. [37]

    Seedream 3.0 technical report.arXiv preprint arXiv:2504.11346, 2025

    Yu Gao, Lixue Gong, Qiushan Guo, Xiaoxia Hou, Zhichao Lai, Fanshi Li, Liang Li, Xiaochen Lian, Chao Liao, Liyang Liu, et al. Seedream 3.0 technical report.arXiv preprint arXiv:2504.11346, 2025. 5, 9

    2025 arXiv

  30. [38]

    Mogao: An omni foundation model for interleaved multi-modal generation

    Chao Liao, Liyang Liu, Xun Wang, Zhengxiong Luo, Xinyu Zhang, Wenliang Zhao, Jie Wu, Liang Li, Zhi Tian, and Weilin Huang. Mogao: An omni foundation model for interleaved multi-modal generation. arXiv preprint arXiv:2505.05472, 2025. 5, 9

    2025 arXiv

  31. [39]

    Ddmi: Domain-agnostic latent diffusion models for synthesizing high-quality implicit neural representations.arXiv preprint arXiv:2401.12517,

    Dogyun Park, Sihyeon Kim, Sojin Lee, and Hyunwoo J Kim. Ddmi: Domain-agnostic latent diffusion models for synthesizing high-quality implicit neural representations.arXiv preprint arXiv:2401.12517,

    arXiv

  32. [40]

    Coco-gan: Generation by parts via conditional coordinating

    Chieh Hubert Lin, Chia-Che Chang, Yu-Sheng Chen, Da-Cheng Juan, Wei Wei, and Hwann-Tzong Chen. Coco-gan: Generation by parts via conditional coordinating. InProceedings of the IEEE/CVF international conference on computer vision, pages 4512–4521, 2019. 5, 12

    2019

  33. [41]

    Image neural field diffusion models

    Yinbo Chen, Oliver Wang, Richard Zhang, Eli Shechtman, Xiaolong Wang, and Michael Gharbi. Image neural field diffusion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8007–8017, 2024. 5, 11

    2024

  34. [42]

    Mip-nerf: A multiscale representation for anti-aliasing neural radiance fields

    Jonathan T Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin-Brualla, and Pratul P Srinivasan. Mip-nerf: A multiscale representation for anti-aliasing neural radiance fields. InProceedings of the IEEE/CVF international conference on computer vision, pages 58...

    2021

  35. [43]

    Neural volumes: Learning dynamic renderable volumes from images.arXiv preprint arXiv:1906.07751,

    Stephen Lombardi, Tomas Simon, Jason Saragih, Gabriel Schwartz, Andreas Lehrmann, and Yaser Sheikh. Neural volumes: Learning dynamic renderable volumes from images.arXiv preprint arXiv:1906.07751,

    1906 arXiv

  36. [44]

    Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction.arXiv preprint arXiv:2106.10689, 2021

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction.arXiv preprint arXiv:2106.10689, 2021. 5

    2021 arXiv

  37. [45]

    Monosdf: Exploring monocular geometric cues for neural implicit surface reconstruction.Advances in neural information processing systems, 35:25018–25032, 2022

    Zehao Yu, Songyou Peng, Michael Niemeyer, Torsten Sattler, and Andreas Geiger. Monosdf: Exploring monocular geometric cues for neural implicit surface reconstruction.Advances in neural information processing systems, 35:25018–25032, 2022. 5

    2022

  38. [46]

    Neus2: Fast learning of neural implicit surfaces for multi-view reconstruction

    Yiming Wang, Qin Han, Marc Habermann, Kostas Daniilidis, Christian Theobalt, and Lingjie Liu. Neus2: Fast learning of neural implicit surfaces for multi-view reconstruction. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 3295–3306, 2023. 5

    2023

  39. [47]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium.Advances in neural information processing systems, 30, 2017

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilibrium.Advances in neural information processing systems, 30, 2017. 6

    2017

  40. [48]

    Generating images with sparse representations.arXiv preprint arXiv:2103.03841, 2021

    Charlie Nash, Jacob Menick, Sander Dieleman, and Peter W Battaglia. Generating images with sparse representations.arXiv preprint arXiv:2103.03841, 2021. 6

    2021 arXiv

  41. [49]

    Improved techniques for training gans.Advances in neural information processing systems, 29, 2016

    Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung, Alec Radford, and Xi Chen. Improved techniques for training gans.Advances in neural information processing systems, 29, 2016. 6

    2016

  42. [50]

    Improved precision and recall metric for assessing generative models.Advances in neural information processing systems, 32,

    Tuomas Kynkäänniemi, Tero Karras, Samuli Laine, Jaakko Lehtinen, and Timo Aila. Improved precision and recall metric for assessing generative models.Advances in neural information processing systems, 32,

  43. [51]

    Geneval: An object-focused framework for evaluating text-to-image alignment.Advances in Neural Information Processing Systems, 36:52132–52152,

    Dhruba Ghosh, Hannaneh Hajishirzi, and Ludwig Schmidt. Geneval: An object-focused framework for evaluating text-to-image alignment.Advances in Neural Information Processing Systems, 36:52132–52152,

  44. [52]

    Ella: Equip diffusion models with llm for enhanced semantic alignment.arXiv preprint arXiv:2403.05135, 2024

    Xiwei Hu, Rui Wang, Yixiao Fang, Bin Fu, Pei Cheng, and Gang Yu. Ella: Equip diffusion models with llm for enhanced semantic alignment.arXiv preprint arXiv:2403.05135, 2024. 6, 11

    2024 arXiv

  45. [53]

    Scaling rectified flow transformers for high-resolution image synthesis.arXiv preprint arXiv:2403.03206, 2024

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling rectified flow transformers for high-resolution image synthesis.arXiv preprint arXiv:2403.03206, 2024. 6, 10

    2024 arXiv

  46. [54]

    Representation alignment for generation: Training diffusion transformers is easier than you think

    Sihyun Yu, Sangkyung Kwak, Huiwon Jang, Jongheon Jeong, Jonathan Huang, Jinwoo Shin, and Saining Xie. Representation alignment for generation: Training diffusion transformers is easier than you think. arXiv preprint arXiv:2410.06940, 2024. 6, 8, 9

    arXiv

  47. [55]

    Dinov2: Learning robust visual features without supervision.arXiv preprint arXiv:2304.07193, 2023

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision.arXiv preprint arXiv:2304.07193, 2023. 6

    2023 arXiv

  48. [56]

    Classifier-free diffusion guidance.arXiv preprint arXiv:2207.12598, 2022

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance.arXiv preprint arXiv:2207.12598, 2022. 8

    2022 arXiv

  49. [57]

    Applying guidance in a limited interval improves sample and distribution quality in diffusion models.arXiv preprint arXiv:2404.07724, 2024

    Tuomas Kynkäänniemi, Miika Aittala, Tero Karras, Samuli Laine, Timo Aila, and Jaakko Lehtinen. Applying guidance in a limited interval improves sample and distribution quality in diffusion models.arXiv preprint arXiv:2404.07724, 2024. 8, 9

    2024 arXiv

  50. [58]

    Autoregressive image generation without vector quantization.Advances in Neural Information Processing Systems, 37:56424–56445, 2025

    Tianhong Li, Yonglong Tian, He Li, Mingyang Deng, and Kaiming He. Autoregressive image generation without vector quantization.Advances in Neural Information Processing Systems, 37:56424–56445, 2025. 8

    2025

  51. [59]

    Causal diffusion transformers for generative modeling.arXiv preprint arXiv:2412.12095, 2024

    Chaorui Deng, Deyao Zh, Kunchang Li, Shi Guan, and Haoqi Fan. Causal diffusion transformers for generative modeling.arXiv preprint arXiv:2412.12095, 2024. 8

    2024 arXiv

  52. [60]

    Jetformer: An autoregressive generative model of raw images and text.arXiv preprint arXiv:2411.19722, 2024

    Michael Tschannen, André Susano Pinto, and Alexander Kolesnikov. Jetformer: An autoregressive generative model of raw images and text.arXiv preprint arXiv:2411.19722, 2024. 8, 9

    2024 arXiv

  53. [61]

    Scalable adaptive computation for iterative generation.arXiv preprint arXiv:2212.11972, 2022

    Allan Jabri, David Fleet, and Ting Chen. Scalable adaptive computation for iterative generation.arXiv preprint arXiv:2212.11972, 2022. 9

    2022 arXiv

  54. [62]

    Show-o: One single transformer to unify multimodal understanding and generation.arXiv preprint arXiv:2408.12528, 2024

    Jinheng Xie, Weijia Mao, Zechen Bai, David Junhao Zhang, Weihao Wang, Kevin Qinghong Lin, Yuchao Gu, Zhijie Chen, Zhenheng Yang, and Mike Zheng Shou. Show-o: One single transformer to unify multimodal understanding and generation.arXiv preprint arXiv:2408.12528, 2024. 10

    2024 arXiv

  55. [63]

    Fluid: Scaling autoregressive text-to-image generative models with continuous tokens.arXiv preprint arXiv:2410.13863, 2024

    Lijie Fan, Tianhong Li, Siyang Qin, Yuanzhen Li, Chen Sun, Michael Rubinstein, Deqing Sun, Kaiming He, and Yonglong Tian. Fluid: Scaling autoregressive text-to-image generative models with continuous tokens.arXiv preprint arXiv:2410.13863, 2024. 9, 10

    arXiv

  56. [64]

    Simplear: Pushing the frontier of autoregressive visual generation through pretraining, sft, and rl.arXiv preprint arXiv:2504.11455, 2025

    Junke Wang, Zhi Tian, Xun Wang, Xinyu Zhang, Weilin Huang, Zuxuan Wu, and Yu-Gang Jiang. Simplear: Pushing the frontier of autoregressive visual generation through pretraining, sft, and rl.arXiv preprint arXiv:2504.11455, 2025. 10

    2025 arXiv

  57. [65]

    Transfusion: Predict the next token and diffuse images with one multi-modal model.arXiv preprint arXiv:2408.11039, 2024

    Chunting Zhou, Lili Yu, Arun Babu, Kushal Tirumala, Michihiro Yasunaga, Leonid Shamis, Jacob Kahn, Xuezhe Ma, Luke Zettlemoyer, and Omer Levy. Transfusion: Predict the next token and diffuse images with one multi-modal model.arXiv preprint arXiv:2408.11039, 2024. 10

    2024 arXiv

  58. [66]

    Segment anything

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alexander C Berg, Wan-Yen Lo, et al. Segment anything. InProceedings of the IEEE/CVF international conference on computer vision, pages 4015–4026, 2023. 9

    2023

  59. [67]

    Journeydb: A benchmark for generative image understanding.Advances in neural information processing systems, 36:49659–49678, 2023

    Keqiang Sun, Junting Pan, Yuying Ge, Hao Li, Haodong Duan, Xiaoshi Wu, Renrui Zhang, Aojun Zhou, Zipeng Qin, Yi Wang, et al. Journeydb: A benchmark for generative image understanding.Advances in neural information processing systems, 36:49659–49678, 2023. 9

    2023

  60. [68]

    Imagenet large scale visual recognition challenge

    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, et al. Imagenet large scale visual recognition challenge. International journal of computer vision, 115:211–252, 2015. 9

    2015

  61. [69]

    Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution.arXiv preprint arXiv:2409.12191, 2024

    Peng Wang, Shuai Bai, Sinan Tan, Shijie Wang, Zhihao Fan, Jinze Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, et al. Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution.arXiv preprint arXiv:2409.12191, 2024. 9, 10

    2024 arXiv

  62. [70]

    Native-resolution image synthesis

    Zidong Wang, Lei Bai, Xiangyu Yue, Wanli Ouyang, and Yiyuan Zhang. Native-resolution image synthesis. arXiv preprint arXiv:2506.03131, 2025. 9 15

    2025 arXiv

  63. [71]

    Blip3-o: A family of fully open unified multimodal models-architecture, training and dataset.arXiv preprint arXiv:2505.09568, 2025

    Jiuhai Chen, Zhiyang Xu, Xichen Pan, Yushi Hu, Can Qin, Tom Goldstein, Lifu Huang, Tianyi Zhou, Saining Xie, Silvio Savarese, et al. Blip3-o: A family of fully open unified multimodal models-architecture, training and dataset.arXiv preprint arXiv:2505.09568, 2025. 9

    2025 arXiv

  64. [72]

    Hyper-sd: Trajectory segmented consistency model for efficient image synthesis.arXiv preprint arXiv:2404.13686,

    Yuxi Ren, Xin Xia, Yanzuo Lu, Jiacheng Zhang, Jie Wu, Pan Xie, Xing Wang, and Xuefeng Xiao. Hyper-sd: Trajectory segmented consistency model for efficient image synthesis.arXiv preprint arXiv:2404.13686,

    arXiv

  65. [73]

    Inrflow: Flow matching for inrs in ambient space.arXiv preprint arXiv:2412.03791, 2024

    Yuyang Wang, Anurag Ranjan, Josh Susskind, and Miguel Angel Bautista. Inrflow: Flow matching for inrs in ambient space.arXiv preprint arXiv:2412.03791, 2024. 11

    2024 arXiv

  66. [74]

    Efficient training with denoised neural weights

    Yifan Gong, Zheng Zhan, Yanyu Li, Yerlan Idelbayev, Andrey Zharkov, Kfir Aberman, Sergey Tulyakov, Yanzhi Wang, and Jian Ren. Efficient training with denoised neural weights. InEuropean Conference on Computer Vision, pages 18–34. Springer, 2024. 11

    2024

  67. [75]

    Patch diffusion: Faster and more data-efficient training of diffusion models

    Zhendong Wang, Yifan Jiang, Huangjie Zheng, Peihao Wang, Pengcheng He, Zhangyang Wang, Weizhu Chen, Mingyuan Zhou, et al. Patch diffusion: Faster and more data-efficient training of diffusion models. Advances in neural information processing systems, 36:72137–72154, 2023. 11

    2023

  68. [76]

    Arbitrary-scale image synthesis

    Evangelos Ntavelis, Mohamad Shahbazi, Iason Kastanis, Radu Timofte, Martin Danelljan, and Luc Van Gool. Arbitrary-scale image synthesis. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11533–11542, 2022. 12

    2022

  69. [77]

    Adversarial generation of continuous images

    Ivan Skorokhodov, Savva Ignatyev, and Mohamed Elhoseiny. Adversarial generation of continuous images. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10753–10764, 2021. 12

    2021

  70. [78]

    Alias-free generative adversarial networks.Advances in neural information processing systems, 34:852–863, 2021

    Tero Karras, Miika Aittala, Samuli Laine, Erik Härkönen, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Alias-free generative adversarial networks.Advances in neural information processing systems, 34:852–863, 2021. 12 16

    2021

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.