REVIEW 5 major objections 4 minor 293 references
XYZFlow:Scaling Multi dimensional Shortcut Flows for Efficient Generative Modeling
T0 review · 5 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash
Pith's one-line read By conditioning every denoising step on the current patch's full history and the complete trajectories of earlier patches, XYZFlow straightens probability flows so that a 172M-parameter model matches or beats one-step models several times…
desk verdict A useful empirical recipe for trajectory-conditioned patch autoregression, with an over-sold theory and no direct test of the straightening mechanism. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is Next Shortcut Prediction, an autoregressive flow-matching distillation objective. It defines a progressive denoising schedule $T(p) = T_\mathrm{full} - \Delta T \cdot (p-1)$ across $P$ patches, so earlier patches anchor the image with full step budgets while later patches take shortcuts. Each predicted next state is produced by an Euler map $G_\theta(x^p_{T(p):t}, t, T_{<p}) = x^p_t + (\gamma(t-1)-\gamma(t)) v_\theta(\cdot)$, trained with a block-wise causal attention mask so the student sees the entire trajectory history and all previous patch trajectories at once. The machinery's role is to turn richer context into a concrete training signal: the regression target is the teacher's next state, and the conditioning is what, in the paper's account, makes that target close to deterministic.
What would settle it
Record a teacher's ODE trajectories and estimate $\mathrm{Var}[x^p_{t-1} \mid x^p_{T(p):t}, T_{<p}]$ by sampling multiple completions from the same conditioning context for early versus late patches. If the variance does not drop sharply for later patches—or if dropping the adversarial loss makes the without-adversarial FID revert to near the unconditioned distillation baseline (about FID 3.0 on the Base model)—then the Dirac-delta premise is unsupported.
Extended reading notes
Core claim
The paper's central claim is that flow matching gains expressivity from structured multidimensional conditioning: temporal scaling conditions each step on the patch's full generation history $x^p_{T(p):t}$, and spatial scaling conditions each patch on the complete trajectories $T_{<p}$ of earlier patches in an autoregressive sequence. Under these conditions the reverse transition $p(x^p_{t-1} \mid x^p_{T(p):t}, T_{<p})$ is argued to approach a Dirac delta, so the one-to-many denoising map becomes one-to-one. Next Shortcut Prediction operationalizes this by giving later patches fewer denoising steps—e.g., a $5 \to 4 \to 3 \to 2$ schedule—and distills the teacher's ODE trajectory with an L2 regression against the teacher's next state. The paper reports this matches the quality of constant-step distillation with 30% fewer total steps and, with optional adversarial fine-tuning, beats teacher and one-step baselines across three model sizes.
Load-bearing premise
The method assumes that conditioning on a patch's full denoising history and on earlier patches' complete trajectories makes each reverse transition $p(x^p_{t-1} \mid x^p_{T(p):t}, T_{<p})$ nearly a Dirac delta; if that transition keeps real multimodality, the L2 distillation target becomes a conditional mean and the reported FID gains would come from the adversarial fine-tuning rather than from flow straightening.
Editorial extensions
If this is right
- Few-step generation can be improved by architectural conditioning rather than by larger teacher models or more elaborate distillation losses.
- A compact 172M-parameter student can match the wall-clock speed of a 676M-parameter one-step model and beat its FID, so model size is not the only axis for quality-latency trade-offs.
- Progressive step reduction ($5 \to 4 \to 3 \to 2$) preserves FID while cutting total steps by 30%, meaning later patches genuinely need fewer denoising iterations.
- The method stays effective when distilling from a weaker 25-step DiT teacher, where ordinary 5-step distillation degrades to FID 8.97 but XYZFlow without the adversarial loss holds 3.85 and with it reaches 1.74.
- Adversarial fine-tuning adds consistent gains on top of the distilled trajectories, indicating the straightened paths give a favorable starting point for high-frequency detail.
Reading between the lines
- The paper treats autoregressive generation as a variance-reduction mechanism; a direct test would measure per-step conditional variance of teacher trajectories and check whether it drops with patch index, as the contraction argument predicts.
- A natural extension the paper leaves untested is applying trajectory conditioning to 3D and audio, where views or frequency bands provide a natural 'patch' ordering and earlier full trajectories could anchor later ones.
- The ablation implying full history matters more than local history suggests the key-value cache design is not an implementation detail; memory-efficient variants of trajectory caching may determine whether the method scales to very long sequences.
- Because the without-adversarial results still trail the teacher on some scales, one could test whether the Dirac-delta premise is better served by learning a stochastic transition rather than relying on adversarial fine-tuning.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes XYZFlow, a patch-autoregressive few-step flow-matching distillation method. Each patch's denoising step is conditioned on the full denoising history of that patch and on the complete trajectories of previously generated patches, and later patches are assigned progressively fewer denoising steps (Next Shortcut Prediction). The authors report ImageNet 256×256 FID scores of 1.63, 1.25, and 1.22 for 172M, 608M, and 1.1B parameter students at 14 total inference steps, with speedups of 36.1×, 9.9×, and 4.0× over the corresponding teachers; they also include weak-teacher experiments and ablations. The paper argues that temporal and spatial conditioning straightens probability flows, and it provides an appendix of theorems intended to support this claim. I agree with the stress-test note that the Dirac-delta objection to Eq. (8) does not land, because the objective regresses against deterministic teacher ODE targets; however, the paper's central mechanistic claim about flow straightening is not quantitatively demonstrated, and the theoretical appendix does not constitute a valid derivation.
Significance. If the empirical results hold, XYZFlow is a practically useful efficient generative modeling method: the weak-teacher check in Table 2 and the component ablations in Tables 4–6 are valuable, and the external FID evaluation makes the core empirical comparison non-circular. The consistently reported quality-latency improvements across model scales are a real strength. At the same time, the paper's distinctive conceptual contribution—that multidimensional conditioning makes flows intrinsically straighter—is not established: the only direct straightness metric, S in Eq. (2), is never reported, and the supporting theorems in Appendix C are asserted rather than proved. The conceptual significance therefore currently rests on an engineering heuristic with good empirical results rather than on the claimed theoretical mechanism.
major comments (5)
- [Section 3.2, Eq. (2) and Figure 2] The central claim of the paper is that conditioning straightens probability flows, and the only direct quantitative evidence for this is the straightness metric S defined in Eq. (2). The text states that "our experiments have demonstrated that S decreases for patches generated later in the sequence," but no quantitative S values are reported anywhere in the paper, and Figure 2 is a schematic illustration rather than a measurement. Because the abstract and introduction attribute the empirical gains to flow straightening, this omission is load-bearing; the ablations in Table 4 show that context helps, but they do not show that trajectories are straighter.
- [Appendix C.4, Theorem C.6] Theorem C.6 claims exponential trajectory alignment, but the proof assumes as its premise the contractive mapping inequality ∥f(v)−f(v′)∥≤α∥v−v′∥ with α<1 (Eq. 24). That contractivity is essentially the exponential-alignment conclusion being proved. This circularity invalidates the theorem and, by extension, Corollaries C.8 and C.9, which rely on the same contraction factors. The proof needs either a derivation of α<1 from the conditioning structure or an explicit statement that contractivity is a separate assumption rather than a proved result.
- [Appendix C.2, Theorem C.4] Theorem C.4 is not a valid derivation. In the proof, the straightness metric is decomposed as if v_θ were a fixed vector, whereas v_θ is a function of x_t; the cross term is said to vanish "due to orthogonality" without specifying the probability space in which this orthogonality holds; and the final quantitative bound ∆S ≥ E[Var[v_θ|H_t]]/(L²T²) is asserted rather than derived. The law of total variance is applied to v_θ rather than to the target (x_1−x_0), so the inequality does not follow from the stated assumptions. This theorem is one of the two main theoretical supports for the straightening claim, so the appendix cannot be cited as a proof.
- [Appendix C.1, Theorem C.2] Theorem C.2's inequality in Eq. (9) is not a consequence of the entropy power inequality or the De Bruijn identity as stated. The proof begins with H(x|c)=H(x)−I(x;c) and then asserts an EPI-based lower bound involving I(x;c)/Var[x], but no argument shows why this bound holds for general distributions, and the role of λ_min(Σ_x) is not explained. Since this theorem underlies the paper's variance-reduction narrative, the information-theoretic foundation needs to be either rigorously established or removed.
- [Section 4.3, Table 4] The ablation labeled "- Full History" removes all inter-patch context, not just the trajectory representation; there is no variant that conditions on the final states of previous patches while omitting their trajectories. Consequently, Table 4 shows that inter-patch context is useful, but it does not support the specific claim in Section 4.3 that "complete trajectory information can provide richer contextual signals than final patch content alone." This distinction is essential to the spatial-scaling contribution and needs a dedicated ablation before the trajectory-transfer mechanism can be credited for the gains.
minor comments (4)
- [Section 5] The concluding remarks contain a typo: "XYZFLow" should be "XYZFlow."
- [Table 1 and Table 4] The colored numbers are described as indicating performance changes, but in the plain-text version the colors are not visible; the direction of each change is already shown by arrows, so the color commentary is redundant and should be removed or replaced with explicit deltas.
- [Appendix A] The discriminator setup says the head architecture is "proposed in the same work" but does not identify which work; a citation is needed.
- [Figure 5] The caption states the samples were "generated by xAR," which appears inconsistent with the surrounding text that attributes the samples to XYZFlow; the caption should be corrected.
Circularity Check
The measured FID and distillation are externally validated, but the appendix's theoretical straightening claims reduce to their own assumptions.
-
self definitional
[Appendix C.2, Theorem C.4, Eqs. (12)-(16)]
"Theorem C.4 (Trajectory Straightening with Quantitative Bounds). ... ΔS = S_unconditional − S_conditional ≥ E[Var[vθ|Ht]]/(L^2T^2). ... For the conditional case, by the law of total variance: Var[v^conditional_θ] = E[Var[v^conditional_θ|Ht]] ≤ E[Var[v^unconditional_θ|Ht]] (by conditioning)."
The theorem is supposed to prove that complete historical trajectory conditioning straightens the flow (reduces S). The proof's only step that establishes the conditional improvement is the parenthetical '(by conditioning)', which asserts that conditioning reduces the variance of the velocity prediction. That assertion is exactly the variance-reduction/straightening phenomenon at issue. No independent bound on E[Var[vθ|Ht]] or on the bias change is derived, so the claimed ΔS bound is the premise restated under a new symbol.
-
self definitional
[Appendix C.4, Theorem C.6, Eqs. (22)-(24)]
"Theorem C.6 (Trajectory Alignment with Exponential Convergence). ... E[∥v_i^t−v_j^t∥^2] ≤ C·e^{−λ(i−j)} + ε_approx. ... The alignment process can be viewed as contractive mapping: v_i^t = f(v^{i−1}_t,H_t)+w_t; ∥f(v)−f(v′)∥≤α∥v−v′∥ with α<1. By contraction mapping principle, the distance between consecutive trajectories decreases geometrically."
The theorem promises exponential decay of the trajectory alignment error as the number of conditioning patches grows. The proof's substantive input is the assumption that the map f is a contraction with α<1; exponential convergence is the definitional consequence of that contraction assumption. The paper never derives α<1 from the temporal or spatial conditioning of XYZFlow, so the exponential-alignment conclusion is identical to the contractivity hypothesis, not a derived result. Corollaries C.8-C.9 inherit the same assumed α_k<1.
full rationale
The empirical core of XYZFlow is not circular: Eq. (8) regresses the student against deterministic ODE states of a fixed teacher, and the FID/IS/Precision/Recall numbers are measured against ImageNet and external baselines, with controlled ablations (Tables 4-6). Removing full history or shortcut prediction changes measured FID, and those observations do not reduce to the loss definition. The circularity is confined to the paper's theoretical apparatus. Theorem C.4 proves 'trajectory straightening with quantitative bounds' by writing the conditional-variance reduction as '(by conditioning)', i.e., assuming the very effect it claims to establish. Theorem C.6 proves exponential trajectory alignment by assuming a contraction constant α<1, which is the exponential-alignment claim in different notation; Corollaries C.8-C.9 then inherit α_k<1 as inputs. Thus the statement that multidimensional conditioning 'straightens' flows is not independently derived; it restates the premise. Because the empirical distillation results and external FID evaluation stand independently, the overall circularity is partial rather than total: score 6.
Assumptions & free parameters
free parameters (4)
- Per-patch step schedule (T_full, Delta_T) =
5 -> 4 -> 3 -> 2
- Patch grid size P =
4 patches (2x2)
- Cell size k =
8x8 latent tokens
- Classifier-free guidance scale =
2.3
assumptions (4)
- domain assumption The reverse transition p(x^p_{t-1} | x^p_{T(p):t}, T_<p) is close to a Dirac delta under strong conditioning.
- domain assumption The actual trained models satisfy L-Lipschitz score, optimal training, and O(Delta_t^2) discretization error.
- ad hoc to paper Trajectory alignment is a contraction with alpha<1.
- domain assumption Precomputed 50-step teacher ODE trajectories with CFG scale 2.3 are an adequate and sufficient distillation target.
Cite this review
Pith. "Pith review of XYZFlow:Scaling Multi dimensional Shortcut Flows for Efficient Generative Modeling." pith.science (2026). https://pith.science/paper/GJ4EED2R
@misc{pith2026260812276,
author = {Pith},
title = {Pith review of: XYZFlow:Scaling Multi dimensional Shortcut Flows for Efficient Generative Modeling},
year = {2026},
howpublished = {\url{https://pith.science/paper/GJ4EED2R}},
note = {Machine review of arXiv:2608.12276}
}
read the original abstract
High-fidelity image generation faces a trade-off between speed and quality. Diffusion models produce strong visuals but require costly iterative sampling. Existing efficient methods mainly distill pretrained models into few-step samplers, a challenging process that depends heavily on teacher-model quality. In this paper, we introduce XYZFlow, a framework that rethinks efficient generation through multidimensional scaling of flow matching. Unlike single-step mappings, XYZFlow enhances expressivity by making probability paths more identifiable and learnable through structured multidimensional conditioning. We view autoregressive modeling as implicit flow straightening, where richer context reduces trajectory ambiguity. XYZFlow realizes this idea through two orthogonal dimensions: temporal scaling, which uses non-Markovian conditioning on the full denoising history; and spatial scaling, enabled by Next Shortcut Prediction, which sequentially generates patches using preceding patches' denoising trajectories as priors. Experiments show that XYZFlow achieves state-of-the-art performance, with 7.2-8.5X teacher speedups and competitive FID, while Next Shortcut Prediction delivers superior quality-latency trade-offs over model scaling or step reduction.
Figures
Figures from the paper (2 more)
Reference graph
Works this paper leans on
-
[1]
ICLR , year=
Sequence level training with recurrent neural networks , author=. ICLR , year=
-
[2]
arXiv preprint arXiv:2303.14389 , year=
MDTv2: Masked Diffusion Transformer is a Strong Image Synthesizer , author=. arXiv preprint arXiv:2303.14389 , year=
-
[3]
arXiv preprint arXiv:2309.16609 , year=
Qwen technical report , author=. arXiv preprint arXiv:2309.16609 , year=
-
[4]
Journal of Machine Learning Research , volume=
Palm: Scaling language modeling with pathways , author=. Journal of Machine Learning Research , volume=
-
[5]
arXiv preprint arXiv:2312.11805 , year=
Gemini: a family of highly capable multimodal models , author=. arXiv preprint arXiv:2312.11805 , year=
-
[6]
arXiv preprint arXiv:2407.21783 , year=
The llama 3 herd of models , author=. arXiv preprint arXiv:2407.21783 , year=
-
[7]
arXiv preprint arXiv:2307.09288 , year=
Llama 2: Open foundation and fine-tuned chat models , author=. arXiv preprint arXiv:2307.09288 , year=
-
[8]
arXiv preprint arXiv:2409.04410 , year=
Open-magvit2: An open-source project toward democratizing auto-regressive visual generation , author=. arXiv preprint arXiv:2409.04410 , year=
Show all 293 references
-
[9]
Learning transferable visual models from natural language supervision , author=
-
[10]
arXiv preprint arXiv:1809.11096 , year=
Large scale GAN training for high fidelity natural image synthesis , author=. arXiv preprint arXiv:1809.11096 , year=
-
[11]
CVPR , year=
Maskgit: Masked generative image transformer , author=. CVPR , year=
-
[12]
CVPR , year=
All are worth words: A vit backbone for diffusion models , author=. CVPR , year=
-
[13]
ECCV , year=
SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformers , author=. ECCV , year=
-
[14]
ICCV , year=
Scalable diffusion models with transformers , author=. ICCV , year=
-
[15]
arXiv preprint arXiv:2201.00273 , year =
Axel Sauer and Katja Schwarz and Andreas Geiger , title =. arXiv preprint arXiv:2201.00273 , year =
-
[16]
Return of Unconditional Generation: A Self-supervised Representation Generation Method , year =
Tianhong Li and Dina Katabi and Kaiming He , journal =. Return of Unconditional Generation: A Self-supervised Representation Generation Method , year =
-
[17]
arXiv preprint arXiv:2505.13447 , year=
Mean flows for one-step generative modeling , author=. arXiv preprint arXiv:2505.13447 , year=
-
[19]
arXiv preprint arXiv:2506.23589 , year=
Transition Matching: Scalable and Flexible Generative Modeling , author=. arXiv preprint arXiv:2506.23589 , year=
-
[20]
CVPR , year=
Autoregressive Distillation of Diffusion Transformers , author=. CVPR , year=
-
[21]
arXiv preprint arXiv:2209.15571 , year=
Building normalizing flows with stochastic interpolants , author=. arXiv preprint arXiv:2209.15571 , year=
-
[22]
ICML , year=
Scaling rectified flow transformers for high-resolution image synthesis , author=. ICML , year=
-
[23]
CVPR , year=
Scaling up gans for text-to-image synthesis , author=. CVPR , year=
-
[24]
arXiv preprint arXiv:1607.06450 , year=
Layer normalization , author=. arXiv preprint arXiv:1607.06450 , year=
-
[25]
NeurIPS , year=
Root mean square layer normalization , author=. NeurIPS , year=
-
[26]
NeurIPS , year=
Laion-5b: An open large-scale dataset for training next generation image-text models , author=. NeurIPS , year=
-
[27]
NeurIPS , year=
Datacomp: In search of the next generation of multimodal datasets , author=. NeurIPS , year=
-
[28]
CVPR , year=
High-resolution image synthesis with latent diffusion models , author=. CVPR , year=
-
[29]
arXiv preprint arXiv:2406.06525 , year=
Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation , author=. arXiv preprint arXiv:2406.06525 , year=
-
[30]
NeurIPS , year=
Autoregressive Image Generation without Vector Quantization , author=. NeurIPS , year=
-
[31]
NeurIPS , year=
Diffusion models beat gans on image synthesis , author=. NeurIPS , year=
-
[32]
Journal of Machine Learning Research , volume=
Cascaded diffusion models for high fidelity image generation , author=. Journal of Machine Learning Research , volume=
-
[33]
Large-dit-imagenet , url=
Alpha-VLLM. Large-dit-imagenet , url=
-
[34]
arXiv preprint arXiv:2206.10789 , year=
Scaling autoregressive models for content-rich text-to-image generation , author=. arXiv preprint arXiv:2206.10789 , year=
-
[35]
arXiv preprint arXiv:1810.04805 , year=
Bert: Pre-training of deep bidirectional transformers for language understanding , author=. arXiv preprint arXiv:1810.04805 , year=
-
[36]
Conditional image generation with pixelcnn decoders , author=
-
[37]
NeurIPS , year=
Visual autoregressive modeling: Scalable image generation via next-scale prediction , author=. NeurIPS , year=
-
[38]
arXiv preprint arXiv:2312.00752 , year=
Mamba: Linear-time sequence modeling with selective state spaces , author=. arXiv preprint arXiv:2312.00752 , year=
-
[39]
arXiv preprint arXiv:2405.21060 , year=
Transformers are SSMs: Generalized models and efficient algorithms through structured state space duality , author=. arXiv preprint arXiv:2405.21060 , year=
-
[40]
arXiv preprint arXiv:2401.09417 , year=
Vision mamba: Efficient visual representation learning with bidirectional state space model , author=. arXiv preprint arXiv:2401.09417 , year=
-
[41]
arXiv preprint arXiv:2406.07537 , year=
Autoregressive Pretraining with Mamba in Vision , author=. arXiv preprint arXiv:2406.07537 , year=
-
[42]
arXiv preprint arXiv:2406.01159 , year=
Dimba: Transformer-Mamba Diffusion Models , author=. arXiv preprint arXiv:2406.01159 , year=
-
[43]
Auto-encoding variational bayes , author=
-
[44]
Neural discrete representation learning , author=
-
[45]
Generating diverse high-fidelity images with vq-vae-2 , author=
-
[46]
Taming transformers for high-resolution image synthesis , author=
-
[47]
Neural computation , volume=
A learning algorithm for continually running fully recurrent neural networks , author=. Neural computation , volume=
-
[48]
arXiv preprint arXiv:2411.00776 , year=
Randomized Autoregressive Visual Generation , author=. arXiv preprint arXiv:2411.00776 , year=
-
[49]
arXiv preprint arXiv:2110.04627 , year=
Vector-quantized image modeling with improved vqgan , author=. arXiv preprint arXiv:2110.04627 , year=
-
[50]
Autoregressive image generation using residual quantization , author=
-
[51]
2018 , howpublished=
Improving Language Understanding by Generative Pre-Training , author=. 2018 , howpublished=
2018
-
[52]
2020 , url=
Language Models are Few-Shot Learners , author=. 2020 , url=
2020
-
[53]
Language Model Beats Diffusion--Tokenizer is Key to Visual Generation , author=
-
[54]
2022 , howpublished =
OpenAI , title =. 2022 , howpublished =
2022
-
[55]
arXiv preprint arXiv:2303.08774 , year=
GPT-4 Technical Report , author=. arXiv preprint arXiv:2303.08774 , year=
-
[56]
Proceedings of the IEEE , volume=
Gradient-based learning applied to document recognition , author=. Proceedings of the IEEE , volume=. 1998 , publisher=
1998
-
[57]
Attention is all you need , author=
-
[58]
arXiv preprint arXiv:2408.12245 , year=
Scalable Autoregressive Image Generation with Mamba , author=. arXiv preprint arXiv:2408.12245 , year=
-
[59]
arXiv preprint arXiv:2111.00396 , year=
Efficiently modeling long sequences with structured state spaces , author=. arXiv preprint arXiv:2111.00396 , year=
-
[60]
Combining recurrent, convolutional, and continuous-time models with linear state space layers , author=
-
[61]
arXiv preprint arXiv:2206.13947 , year=
Long range language modeling via gated state spaces , author=. arXiv preprint arXiv:2206.13947 , year=
-
[62]
arXiv preprint arXiv:2206.12037 , year=
How to train your hippo: State space models with generalized orthogonal basis projections , author=. arXiv preprint arXiv:2206.12037 , year=
-
[63]
An Image is Worth 32 Tokens for Reconstruction and Generation , author=
-
[64]
arXiv preprint arXiv:2409.16211 , year=
MaskBit: Embedding-free Image Generation via Bit Tokens , author=. arXiv preprint arXiv:2409.16211 , year=
-
[65]
Alleviating Distortion in Image Generation via Multi-Resolution Diffusion Models , author=
-
[66]
arXiv preprint arXiv:2412.15119 , year=
Parallelized Autoregressive Visual Generation , author=. arXiv preprint arXiv:2412.15119 , year=
-
[67]
and Ermon, Stefano and Rudra, Atri and R
Dao, Tri and Fu, Daniel Y. and Ermon, Stefano and Rudra, Atri and R. Flash
-
[68]
Dao, Tri , booktitle=ICLR, year=. Flash
-
[69]
arXiv preprint arXiv:2302.13971 , year=
Llama: Open and efficient foundation language models , author=. arXiv preprint arXiv:2302.13971 , year=
-
[70]
International Conference on Artificial Intelligence and Statistics , year=
Variational rejection sampling , author=. International Conference on Artificial Intelligence and Statistics , year=
-
[71]
arXiv preprint arXiv:1810.06758 , year=
Discriminator rejection sampling , author=. arXiv preprint arXiv:1810.06758 , year=
-
[72]
simple diffusion: End-to-end diffusion for high resolution images , author=
-
[73]
NeurIPS , year=
Denoising diffusion probabilistic models , author=. NeurIPS , year=
-
[74]
arXiv preprint arXiv:2010.02502 , year=
Denoising diffusion implicit models , author=. arXiv preprint arXiv:2010.02502 , year=
2010 arXiv
-
[75]
Generative modeling by estimating gradients of the data distribution , author=
-
[76]
ICML , year=
Generative Pretraining from Pixels , author=. ICML , year=
-
[77]
Generative adversarial nets , author=
-
[78]
arXiv preprint arXiv:2409.16280 , year=
Monoformer: One transformer for both diffusion and autoregression , author=. arXiv preprint arXiv:2409.16280 , year=
-
[79]
ICLR , year=
Dart: Denoising autoregressive transformer for scalable text-to-image generation , author=. ICLR , year=
-
[80]
arXiv preprint arXiv:2412.12095 , year=
Causal Diffusion Transformers for Generative Modeling , author=. arXiv preprint arXiv:2412.12095 , year=
-
[81]
ECCV , year=
Microsoft coco: Common objects in context , author=. ECCV , year=
-
[82]
MICCAI , year=
U-net: Convolutional networks for biomedical image segmentation , author=. MICCAI , year=
-
[83]
arXiv preprint arXiv:2410.10733 , year=
Deep Compression Autoencoder for Efficient High-Resolution Diffusion Models , author=. arXiv preprint arXiv:2410.10733 , year=
-
[84]
Proceedings of Machine Learning and Systems , volume=
Efficiently scaling transformer inference , author=. Proceedings of Machine Learning and Systems , volume=
-
[85]
2002 , publisher=
Ordinary differential equations , author=. 2002 , publisher=
2002
-
[86]
NeurIPS , year=
Gans trained by a two time-scale update rule converge to a local nash equilibrium , author=. NeurIPS , year=
-
[87]
arXiv preprint arXiv:2010.16061 , year=
Evaluation: from precision, recall and F-measure to ROC, informedness, markedness and correlation , author=. arXiv preprint arXiv:2010.16061 , year=
2010 arXiv
-
[88]
Adam: A method for stochastic optimization , author=
-
[89]
Decoupled weight decay regularization , author=
-
[90]
arXiv preprint arXiv:2410.19324 , year=
Simpler Diffusion (SiD2): 1.5 FID on ImageNet512 with pixel-space diffusion , author=. arXiv preprint arXiv:2410.19324 , year=
-
[91]
arXiv preprint arXiv:2410.06940 , year=
Representation alignment for generation: Training diffusion transformers is easier than you think , author=. arXiv preprint arXiv:2410.06940 , year=
-
[92]
Generation: Taming Optimization Dilemma in Latent Diffusion Models , author=
Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models , author=. arXiv preprint arXiv:2501.01423 , year=
-
[93]
arXiv preprint arXiv:2404.07724 , year=
Applying guidance in a limited interval improves sample and distribution quality in diffusion models , author=. arXiv preprint arXiv:2404.07724 , year=
-
[94]
ICCV , year=
Emerging properties in self-supervised vision transformers , author=. ICCV , year=
-
[95]
arXiv preprint arXiv:2304.07193 , year=
Dinov2: Learning robust visual features without supervision , author=. arXiv preprint arXiv:2304.07193 , year=
-
[96]
ICML , year=
Pixel recurrent neural networks , author=. ICML , year=
-
[97]
CVPR , year=
Attngan: Fine-grained text to image generation with attentional generative adversarial networks , author=. CVPR , year=
-
[98]
CVPR , year=
Dm-gan: Dynamic memory generative adversarial networks for text-to-image synthesis , author=. CVPR , year=
-
[99]
CVPR , year=
Vector quantized diffusion model for text-to-image synthesis , author=. CVPR , year=
-
[100]
CVPR , year=
Df-gan: A simple and effective baseline for text-to-image synthesis , author=. CVPR , year=
-
[101]
CVPR , year=
Cross-modal contrastive learning for text-to-image generation , author=. CVPR , year=
-
[102]
AAAI , year=
Frido: Feature pyramid diffusion for complex scene image synthesis , author=. AAAI , year=
-
[103]
CVPR , year=
Towards language-free training for text-to-image generation , author=. CVPR , year=
-
[104]
arXiv preprint arXiv:2407.01392 , year=
Diffusion forcing: Next-token prediction meets full-sequence diffusion , author=. arXiv preprint arXiv:2407.01392 , year=
-
[105]
ICML , year=
Learning transferable visual models from natural language supervision , author=. ICML , year=
-
[106]
Findings of the Association for Computational Linguistics: ACL 2022 , year=
Why Exposure Bias Matters: An Imitation Learning Perspective of Error Accumulation in Language Generation , author=. Findings of the Association for Computational Linguistics: ACL 2022 , year=
2022
-
[107]
EMNLP , year=
Exposure Bias versus Self-Recovery: Are Distortions Really Incremental for Autoregressive Text Generation? , author=. EMNLP , year=
-
[108]
FirstName LastName , title =
-
[109]
FirstName Alpher , title =
-
[110]
Journal of Foo , volume = 13, number = 1, pages =
FirstName Alpher and FirstName Fotheringham-Smythe , title =. Journal of Foo , volume = 13, number = 1, pages =
-
[111]
Journal of Foo , volume = 14, number = 1, pages =
FirstName Alpher and FirstName Fotheringham-Smythe and FirstName Gamow , title =. Journal of Foo , volume = 14, number = 1, pages =
-
[112]
FirstName Alpher and FirstName Gamow , title =
-
[113]
International Conference on Learning Representations , year=
Score-Based Generative Modeling through Stochastic Differential Equations , author=. International Conference on Learning Representations , year=
-
[114]
arXiv preprint arXiv:2309.15807 , year=
Emu: Enhancing image generation models using photogenic needles in a haystack , author=. arXiv preprint arXiv:2309.15807 , year=
-
[115]
SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis , author=
-
[116]
Forty-first International Conference on Machine Learning , year=
Scaling rectified flow transformers for high-resolution image synthesis , author=. Forty-first International Conference on Machine Learning , year=
-
[117]
The Twelfth International Conference on Learning Representations , year=
PixArt- : Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis , author=. The Twelfth International Conference on Learning Representations , year=
-
[118]
International conference on machine learning , pages=
Deep unsupervised learning using nonequilibrium thermodynamics , author=. International conference on machine learning , pages=. 2015 , organization=
2015
-
[119]
Advances in neural information processing systems , volume=
Generative modeling by estimating gradients of the data distribution , author=. Advances in neural information processing systems , volume=
-
[120]
International Conference on Learning Representations , year=
Denoising Diffusion Implicit Models , author=. International Conference on Learning Representations , year=
-
[121]
ICLR , year=
Progressive Distillation for Fast Sampling of Diffusion Models , author=. ICLR , year=
-
[122]
arXiv preprint arXiv:2101.02388 , year=
Knowledge distillation in iterative generative models for improved sampling speed , author=. arXiv preprint arXiv:2101.02388 , year=
-
[123]
Consistency Trajectory Models: Learning Probability Flow
Dongjun Kim and Chieh-Hsin Lai and Wei-Hsiang Liao and Naoki Murata and Yuhta Takida and Toshimitsu Uesaka and Yutong He and Yuki Mitsufuji and Stefano Ermon , booktitle=. Consistency Trajectory Models: Learning Probability Flow. 2024 , url=
2024
-
[124]
Stochastic Processes and their Applications , volume=
Reverse-time diffusion equation models , author=. Stochastic Processes and their Applications , volume=. 1982 , publisher=
1982
-
[125]
Diffusion schr
De Bortoli, Valentin and Thornton, James and Heng, Jeremy and Doucet, Arnaud , journal=. Diffusion schr
-
[126]
arXiv preprint arXiv:2402.13929 , year=
Sdxl-lightning: Progressive adversarial diffusion distillation , author=. arXiv preprint arXiv:2402.13929 , year=
-
[127]
arXiv preprint arXiv:2311.05556 , year=
Lcm-lora: A universal stable-diffusion acceleration module , author=. arXiv preprint arXiv:2311.05556 , year=
-
[128]
arXiv preprint arXiv:2401.05252 , year=
Pixart- \ delta \ : Fast and controllable image generation with latent consistency models , author=. arXiv preprint arXiv:2401.05252 , year=
-
[129]
arXiv preprint arXiv:2403.04692 , year=
Pixart- sigma: Weak-to-strong training of diffusion transformer for 4k text-to-image generation , author=. arXiv preprint arXiv:2403.04692 , year=
-
[130]
ACM SIGGRAPH 2022 conference proceedings , pages=
Stylegan-xl: Scaling stylegan to large diverse datasets , author=. ACM SIGGRAPH 2022 conference proceedings , pages=
2022
-
[131]
International conference on machine learning , pages=
Pixel recurrent neural networks , author=. International conference on machine learning , pages=. 2016 , organization=
2016
-
[132]
Advances in neural information processing systems , volume=
Conditional image generation with pixelcnn decoders , author=. Advances in neural information processing systems , volume=
-
[133]
European Conference on Computer Vision (ECCV) , year =
Kang, Minguk and Zhang, Richard and Barnes, Connelly and Paris, Sylvain and Kwak, Suha and Park, Jaesik and Shechtman, Eli and Zhu, Jun-Yan and Park, Taesung , title =. European Conference on Computer Vision (ECCV) , year =
-
[134]
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
Generative multimodal models are in-context learners , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
-
[135]
arXiv preprint arXiv:2409.18869 , year=
Emu3: Next-token prediction is all you need , author=. arXiv preprint arXiv:2409.18869 , year=
-
[136]
The Twelfth International Conference on Learning Representations , year=
Emu: Generative Pretraining in Multimodality , author=. The Twelfth International Conference on Learning Representations , year=
-
[137]
2024 , eprint=
Fast High-Resolution Image Synthesis with Latent Adversarial Diffusion Distillation , author=. 2024 , eprint=
2024
-
[138]
Large Scale
Andrew Brock and Jeff Donahue and Karen Simonyan , booktitle=. Large Scale. 2019 , url=
2019
-
[139]
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
Maskgit: Masked generative image transformer , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
-
[140]
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
Revisiting non-autoregressive transformers for efficient image synthesis , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
-
[141]
Transactions on Machine Learning Research , issn=
Scaling Autoregressive Models for Content-Rich Text-to-Image Generation , author=. Transactions on Machine Learning Research , issn=. 2022 , url=
2022
-
[142]
2023 , eprint=
MAGE: MAsked Generative Encoder to Unify Representation Learning and Image Synthesis , author=. 2023 , eprint=
2023
-
[143]
BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding
Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina. BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding. Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Hum...
2019 doi
-
[144]
Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
Taming transformers for high-resolution image synthesis , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
-
[145]
Advances in neural information processing systems , volume=
Gans trained by a two time-scale update rule converge to a local nash equilibrium , author=. Advances in neural information processing systems , volume=
-
[146]
NeurIPS , year=
Improved techniques for training gans , author=. NeurIPS , year=
-
[147]
Advances in neural information processing systems , volume=
Improved precision and recall metric for assessing generative models , author=. Advances in neural information processing systems , volume=
-
[148]
ECCV , year=
Structural Causal 3D Reconstruction , author=. ECCV , year=
-
[149]
ICCV , year=
Self-Supervised 3D Face Reconstruction via Conditional Estimation , author=. ICCV , year=
-
[150]
Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
Going deeper with convolutions , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=
-
[151]
Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 , pages=
Microsoft coco: Common objects in context , author=. Computer Vision--ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 , pages=. 2014 , organization=
2014
-
[152]
Advances in Neural Information Processing Systems , volume=
T2i-compbench: A comprehensive benchmark for open-world compositional text-to-image generation , author=. Advances in Neural Information Processing Systems , volume=
-
[153]
arXiv preprint arXiv:2207.12598 , year=
Classifier-free diffusion guidance , author=. arXiv preprint arXiv:2207.12598 , year=
-
[154]
arXiv preprint arXiv:2310.04378 , year=
Latent consistency models: Synthesizing high-resolution images with few-step inference , author=. arXiv preprint arXiv:2310.04378 , year=
-
[155]
European Conference on Computer Vision , pages=
Adversarial diffusion distillation , author=. European Conference on Computer Vision , pages=. 2024 , organization=
2024
-
[156]
arXiv preprint arXiv:2405.05224 , year=
Imagine flash: Accelerating emu diffusion models with backward distillation , author=. arXiv preprint arXiv:2405.05224 , year=
-
[157]
Advances in neural information processing systems , volume=
Generative adversarial nets , author=. Advances in neural information processing systems , volume=
-
[158]
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
Scaling up gans for text-to-image synthesis , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
-
[159]
Advances in Neural Information Processing Systems , volume=
Understanding diffusion objectives as the elbo with simple data augmentation , author=. Advances in Neural Information Processing Systems , volume=
-
[160]
CVPR , year=
Autoregressive image generation using residual quantization , author=. CVPR , year=
-
[161]
Advances in neural information processing systems , volume=
Generating diverse high-fidelity images with vq-vae-2 , author=. Advances in neural information processing systems , volume=
-
[162]
arXiv preprint arXiv:2312.03701 , year=
Self-conditioned image generation via generating representations , author=. arXiv preprint arXiv:2312.03701 , year=
-
[163]
2022 , url=
Chenlin Meng and Yutong He and Yang Song and Jiaming Song and Jiajun Wu and Jun-Yan Zhu and Stefano Ermon , booktitle=. 2022 , url=
2022
-
[164]
The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
Return of Unconditional Generation: A Self-supervised Representation Generation Method , author=. The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
-
[165]
The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
Autoregressive Image Generation without Vector Quantization , author=. The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
-
[166]
Neural computation , volume=
A connection between score matching and denoising autoencoders , author=. Neural computation , volume=. 2011 , publisher=
2011
-
[167]
Proceedings of the fourteenth international conference on artificial intelligence and statistics , pages=
The neural autoregressive distribution estimator , author=. Proceedings of the fourteenth international conference on artificial intelligence and statistics , pages=. 2011 , organization=
2011
-
[168]
4th International Conference on Learning Representations, ICLR 2016 , year=
Sequence level training with recurrent neural networks , author=. 4th International Conference on Learning Representations, ICLR 2016 , year=
2016
-
[169]
Proceedings of the AAAI conference on artificial intelligence , volume=
Film: Visual reasoning with a general conditioning layer , author=. Proceedings of the AAAI conference on artificial intelligence , volume=
-
[170]
The Twelfth International Conference on Learning Representations , year=
Elucidating the Exposure Bias in Diffusion Models , author=. The Twelfth International Conference on Learning Representations , year=
-
[171]
European Conference on Computer Vision , pages=
Diffit: Diffusion vision transformers for image generation , author=. European Conference on Computer Vision , pages=. 2024 , organization=
2024
-
[172]
arXiv preprint arXiv:2307.01952 , year=
Sdxl: Improving latent diffusion models for high-resolution image synthesis , author=. arXiv preprint arXiv:2307.01952 , year=
-
[173]
The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction , author=. The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
-
[174]
The Twelfth International Conference on Learning Representations , year=
Dustin Podell and Zion English and Kyle Lacey and Andreas Blattmann and Tim Dockhorn and Jonas M. The Twelfth International Conference on Learning Representations , year=
-
[175]
Forty-first International Conference on Machine Learning , year=
Data-free Distillation of Diffusion Models with Bootstrapping , author=. Forty-first International Conference on Machine Learning , year=
-
[176]
International conference on machine learning , pages=
Fast sampling of diffusion models via operator learning , author=. International conference on machine learning , pages=. 2023 , organization=
2023
-
[177]
The Eleventh International Conference on Learning Representations , year=
Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow , author=. The Eleventh International Conference on Learning Representations , year=
-
[178]
ICLR , year=
Streaming Autoregressive Video Generation via Diagonal Distillation , author=. ICLR , year=
-
[179]
arXiv preprint arXiv:2303.04248 , year=
Tract: Denoising diffusion models with transitive closure time-distillation , author=. arXiv preprint arXiv:2303.04248 , year=
-
[180]
Advances in Neural Information Processing Systems , volume=
Diff-instruct: A universal approach for transferring knowledge from pre-trained diffusion models , author=. Advances in Neural Information Processing Systems , volume=
-
[181]
2024 , url=
Sirui Xie and Zhisheng Xiao and Diederik P Kingma and Tingbo Hou and Ying Nian Wu and Kevin Patrick Murphy and Tim Salimans and Ben Poole and Ruiqi Gao , booktitle=. 2024 , url=
2024
-
[182]
The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
One-Step Diffusion Distillation through Score Implicit Matching , author=. The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
-
[183]
arXiv , year=
Multistep consistency models , author=. arXiv , year=
-
[184]
and Park, Taesung , title =
Yin, Tianwei and Gharbi, Micha\"el and Zhang, Richard and Shechtman, Eli and Durand, Fr\'edo and Freeman, William T. and Park, Taesung , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2024 , pages =
2024
-
[185]
Advances in Neural Information Processing Systems , year=
Attention is all you need , author=. Advances in Neural Information Processing Systems , year=
-
[186]
The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
Improved Distribution Matching Distillation for Fast Image Synthesis , author=. The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
-
[187]
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
On distillation of guided diffusion models , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
-
[188]
The Eleventh International Conference on Learning Representations , year=
Generative Modelling with Inverse Heat Dissipation , author=. The Eleventh International Conference on Learning Representations , year=
-
[189]
Dongjun Kim and Chieh-Hsin Lai and Wei-Hsiang Liao and Yuhta Takida and Naoki Murata and Toshimitsu Uesaka and Yuki Mitsufuji and Stefano Ermon , booktitle=. PaGo. 2024 , url=
2024
-
[190]
Dieleman, Sander , title =
-
[191]
Advances in Neural Information Processing Systems , volume=
Projected gans converge faster , author=. Advances in Neural Information Processing Systems , volume=
-
[192]
International conference on machine learning , pages=
Stylegan-t: Unlocking the power of gans for fast large-scale text-to-image synthesis , author=. International conference on machine learning , pages=. 2023 , organization=
2023
-
[193]
NeurIPS , year=
Elucidating the Design Space of Diffusion-Based Generative Models , author=. NeurIPS , year=
-
[194]
2022 , url=
Tim Dockhorn and Arash Vahdat and Karsten Kreis , booktitle=. 2022 , url=
2022
-
[195]
The Eleventh International Conference on Learning Representations , year=
Fast Sampling of Diffusion Models with Exponential Integrator , author=. The Eleventh International Conference on Learning Representations , year=
-
[196]
ICLR , year=
Flow Matching for Generative Modeling , author=. ICLR , year=
-
[197]
The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
Improving the Training of Rectified Flows , author=. The Thirty-eighth Annual Conference on Neural Information Processing Systems , year=
-
[198]
Minimizing Trajectory Curvature of
Lee, Sangyun and Kim, Beomsu and Ye, Jong Chul , booktitle =. Minimizing Trajectory Curvature of. 2023 , editor =
2023
-
[199]
The Twelfth International Conference on Learning Representations , year=
InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generation , author=. The Twelfth International Conference on Learning Representations , year=
-
[200]
NeurIPS , year=
Multistep Distillation of Diffusion Models via Moment Matching , author=. NeurIPS , year=
-
[201]
Advances in Neural Information Processing Systems , volume=
Geneval: An object-focused framework for evaluating text-to-image alignment , author=. Advances in Neural Information Processing Systems , volume=
-
[202]
arXiv preprint arXiv:2412.04431 , year=
Infinity: Scaling bitwise autoregressive modeling for high-resolution image synthesis , author=. arXiv preprint arXiv:2412.04431 , year=
-
[203]
arXiv preprint arXiv:2504.18391 , year=
Fast autoregressive models for continuous latent generation , author=. arXiv preprint arXiv:2504.18391 , year=
-
[204]
Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
Masked autoencoders are scalable vision learners , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
-
[205]
SIAM review , volume=
An algorithmic introduction to numerical simulation of stochastic differential equations , author=. SIAM review , volume=. 2001 , publisher=
2001
-
[206]
Proceedings of the IEEE/CVF winter conference on applications of computer vision , pages=
Common diffusion noise schedules and sample steps are flawed , author=. Proceedings of the IEEE/CVF winter conference on applications of computer vision , pages=
-
[207]
arXiv preprint arXiv:2202.09778 , year=
Pseudo numerical methods for diffusion models on manifolds , author=. arXiv preprint arXiv:2202.09778 , year=
-
[208]
arXiv preprint arXiv:2209.03003 , year=
Flow straight and fast: Learning to generate and transfer data with rectified flow , author=. arXiv preprint arXiv:2209.03003 , year=
-
[209]
arXiv preprint arXiv:2505.07071 , year=
Semantic-guided diffusion model for single-step image super-resolution , author=. arXiv preprint arXiv:2505.07071 , year=
-
[210]
NeurIPS , year=
Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps , author=. NeurIPS , year=
-
[211]
arXiv preprint arXiv:2211.01095 , year=
Dpm-solver++: Fast solver for guided sampling of diffusion probabilistic models , author=. arXiv preprint arXiv:2211.01095 , year=
-
[212]
ECCV , year=
Sit: Exploring flow and diffusion-based generative models with scalable interpolant transformers , author=. ECCV , year=
-
[213]
International conference on machine learning , pages=
Improved denoising diffusion probabilistic models , author=. International conference on machine learning , pages=. 2021 , organization=
2021
-
[214]
Scalable diffusion models with
Peebles, William and Xie, Saining , booktitle=. Scalable diffusion models with
-
[215]
arXiv preprint arXiv:2502.20388 , year=
Beyond next-token: Next-x prediction for autoregressive visual generation , author=. arXiv preprint arXiv:2502.20388 , year=
-
[216]
arXiv preprint arXiv:2011.13456 , year=
Score-based generative modeling through stochastic differential equations , author=. arXiv preprint arXiv:2011.13456 , year=
2011 arXiv
-
[217]
arXiv preprint arXiv:2501.02576 , year=
Depthmaster: Taming diffusion models for monocular depth estimation , author=. arXiv preprint arXiv:2501.02576 , year=
-
[218]
Tschannen, Michael and Eastwood, Cian and Mentzer, Fabian , journal=
-
[219]
2022 , howpublished=
Diffusers: State-of-the-art diffusion models , author=. 2022 , howpublished=
2022
-
[220]
arXiv preprint arXiv:2411.11925 , year=
Continuous speculative decoding for autoregressive image generation , author=. arXiv preprint arXiv:2411.11925 , year=
-
[221]
arXiv preprint arXiv:2503.21979 , year=
Harmonizing visual representations for unified multimodal understanding and generation , author=. arXiv preprint arXiv:2503.21979 , year=
-
[222]
arXiv preprint arXiv:2503.12450 , year=
Lazymar: Accelerating masked autoregressive models via feature caching , author=. arXiv preprint arXiv:2503.12450 , year=
-
[223]
arXiv preprint arXiv:2503.02883 , year=
Arinar: Bi-level autoregressive feature-by-feature generative models , author=. arXiv preprint arXiv:2503.02883 , year=
-
[224]
arXiv preprint arXiv:2410.03080 , year=
Generative edge detection with stable diffusion , author=. arXiv preprint arXiv:2410.03080 , year=
-
[225]
2018 , url=
JAX: composable transformations of Python+NumPy programs , author=. 2018 , url=
2018
-
[226]
International Conference on Learning Representations (ICLR) , year=
Large scale GAN training for high fidelity natural image synthesis , author=. International Conference on Learning Representations (ICLR) , year=
-
[227]
IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year=
Maskgit: Masked generative image transformer , author=. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year=
-
[228]
CVPR , year=
Imagenet: A large-scale hierarchical image database , author=. CVPR , year=
-
[229]
International Conference on Learning Representations (ICLR) , year=
An image is worth 16x16 words: Transformers for image recognition at scale , author=. International Conference on Learning Representations (ICLR) , year=
-
[230]
IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year=
Taming transformers for high-resolution image synthesis , author=. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year=
-
[231]
Cambridge Machine Learning Group Blog , year=
An introduction to flow matching , author=. Cambridge Machine Learning Group Blog , year=
-
[232]
ICLR , year=
One step diffusion via shortcut models , author=. ICLR , year=
-
[233]
arXiv preprint arXiv:2406.14548 , year=
Consistency models made easy , author=. arXiv preprint arXiv:2406.14548 , year=
-
[234]
arXiv preprint arXiv:1706.02677 , year=
Accurate, large minibatch sgd: Training imagenet in 1 hour , author=. arXiv preprint arXiv:1706.02677 , year=
-
[235]
Neural Information Processing Systems (NeurIPS) , year=
GANs trained by a two time-scale update rule converge to a local nash equilibrium , author=. Neural Information Processing Systems (NeurIPS) , year=
-
[236]
International Conference on Machine Learning (ICML) , year=
Simple diffusion: End-to-end diffusion for high resolution images , author=. International Conference on Machine Learning (ICML) , year=
-
[237]
IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year=
Scaling up gans for text-to-image synthesis , author=. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year=
-
[238]
International Conference on Learning Representations (ICLR) , year=
Consistency trajectory models: Learning probability flow ODE trajectory of diffusion , author=. International Conference on Learning Representations (ICLR) , year=
-
[239]
International Conference on Learning Representations (ICLR) , year=
Adam: A method for stochastic optimization , author=. International Conference on Learning Representations (ICLR) , year=
-
[240]
2009 , url=
Learning multiple layers of features from tiny images , author=. 2009 , url=
2009
-
[241]
Virorum celeberrimorum G. G. Leibnitii et Johannis Bernoullii Commercium philosophicum et mathematicum , author=. 1745 , publisher=
-
[242]
Neural Information Processing Systems (NeurIPS) , year=
Autoregressive image generation without vector quantization , author=. Neural Information Processing Systems (NeurIPS) , year=
-
[243]
ICLR , year=
Flow matching for generative modeling , author=. ICLR , year=
-
[244]
2024 , url=
Flow matching guide and code , author=. 2024 , url=
2024
-
[245]
International Conference on Learning Representations (ICLR) , year=
Flow straight and fast: Learning to generate and transfer data with rectified flow , author=. International Conference on Learning Representations (ICLR) , year=
-
[246]
NeurIPS , year=
Diff-instruct: A universal approach for transferring knowledge from pre-trained diffusion models , author=. NeurIPS , year=
-
[247]
Movie Gen: A cast of media foundation models , author=
-
[248]
ICML , year=
Variational inference with normalizing flows , author=. ICML , year=
-
[249]
SIGGRAPH , year=
Stylegan-xl: Scaling stylegan to large diverse datasets , author=. SIGGRAPH , year=
-
[250]
ECCV , year=
Adversarial diffusion distillation , author=. ECCV , year=
-
[251]
ICML , year=
Deep unsupervised learning using nonequilibrium thermodynamics , author=. ICML , year=
-
[252]
ICLR , year=
Improved techniques for training consistency models , author=. ICLR , year=
-
[253]
NeurIPS , year=
Generative modeling by estimating gradients of the data distribution , author=. NeurIPS , year=
-
[254]
ICLR , year=
Score-based generative modeling through stochastic differential equations , author=. ICLR , year=
-
[255]
ICML , year=
Consistency models , author=. ICML , year=
-
[256]
Neural Information Processing Systems (NeurIPS) , year=
Attention is all you need , author=. Neural Information Processing Systems (NeurIPS) , year=
-
[257]
CVPR , year=
One-step diffusion with distribution matching distillation , author=. CVPR , year=
-
[258]
International Conference on Learning Representations (ICLR) , year=
Representation alignment for generation: Training diffusion transformers is easier than you think , author=. International Conference on Learning Representations (ICLR) , year=
-
[259]
arXiv preprint arXiv:2503.07565 , year=
Inductive moment matching , author=. arXiv preprint arXiv:2503.07565 , year=
-
[260]
International Conference on Machine Learning (ICML) , year=
Score identity distillation: Exponentially fast distillation of pretrained diffusion models for one-step generation , author=. International Conference on Machine Learning (ICML) , year=
-
[261]
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
Taming transformers for high-resolution image synthesis , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
-
[262]
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
Masked autoencoders are scalable vision learners , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
-
[263]
Advances in Neural Information Processing Systems , volume=
GANs trained by a two time-scale update rule converge to a local nash equilibrium , author=. Advances in Neural Information Processing Systems , volume=
-
[264]
SIAM Review , volume=
An algorithmic introduction to numerical simulation of stochastic differential equations , author=. SIAM Review , volume=
-
[265]
Advances in Neural Information Processing Systems , volume=
Autoregressive image generation without vector quantization , author=. Advances in Neural Information Processing Systems , volume=
-
[266]
Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages=
Common diffusion noise schedules and sample steps are flawed , author=. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages=
-
[267]
European Conference on Computer Vision , pages=
Sit: Exploring flow and diffusion-based generative models with scalable interpolant transformers , author=. European Conference on Computer Vision , pages=. 2024 , publisher=
2024
-
[268]
International Conference on Machine Learning , pages=
Improved denoising diffusion probabilistic models , author=. International Conference on Machine Learning , pages=. 2021 , publisher=
2021
-
[269]
Advances in Neural Information Processing Systems , volume=
Generating diverse high-fidelity images with VQ-VAE-2 , author=. Advances in Neural Information Processing Systems , volume=
-
[270]
arXiv preprint arXiv:2412.15205 , year=
Flowar: Scale-wise autoregressive image generation meets flow matching , author=. arXiv preprint arXiv:2412.15205 , year=
-
[271]
Advances in Neural Information Processing Systems , volume=
Improved techniques for training GANs , author=. Advances in Neural Information Processing Systems , volume=
-
[272]
ICLR , year=
Denoising diffusion implicit models , author=. ICLR , year=
-
[273]
Advances in Neural Information Processing Systems , volume=
Generative modeling by estimating gradients of the data distribution , author=. Advances in Neural Information Processing Systems , volume=
-
[274]
arXiv preprint arXiv:2312.02116 , year=
GIVT: Generative infinite-vocabulary Transformers , author=. arXiv preprint arXiv:2312.02116 , year=
-
[275]
2022 , url=
Diffusers: State-of-the-art diffusion models , author=. 2022 , url=
2022
-
[276]
arXiv preprint arXiv:2406.07507 , year=
Flow map matching , author=. arXiv preprint arXiv:2406.07507 , year=
-
[277]
2024 , url=
An introduction to flow matching , author=. 2024 , url=
2024
-
[278]
ICLR , year=
Flow straight and fast: Learning to generate and transfer data with rectified flow , author=. ICLR , year=
-
[279]
European Conference on Computer Vision (ECCV) , year=
Sit: Exploring flow and diffusion-based generative models with scalable interpolant transformers , author=. European Conference on Computer Vision (ECCV) , year=
-
[280]
International Conference on Machine Learning (ICML) , year=
Variational inference with normalizing flows , author=. International Conference on Machine Learning (ICML) , year=
-
[281]
Medical Image Computing and Computer-Assisted Intervention (MICCAI) , year=
U-net: Convolutional networks for biomedical image segmentation , author=. Medical Image Computing and Computer-Assisted Intervention (MICCAI) , year=
-
[282]
Stylegan-xl: Scaling stylegan to large diverse datasets , author=
-
[283]
ICLR , year=
Building normalizing flows with stochastic interpolants , author=. ICLR , year=
-
[284]
Forty-first international conference on machine learning , year=
Scaling rectified flow transformers for high-resolution image synthesis , author=. Forty-first international conference on machine learning , year=
-
[285]
2024 , month=
An introduction to flow matching , author=. 2024 , month=
2024
-
[286]
NeurIPS , year=
One-step diffusion distillation via deep equilibrium models , author=. NeurIPS , year=
-
[287]
Neural Information Processing Systems (NeurIPS) , year=
Elucidating the design space of diffusion-based generative models , author=. Neural Information Processing Systems (NeurIPS) , year=
-
[288]
Virorum celeberrimorum G
Epistola LXXI ad johannem bernoullium, 5 aug 1697 , author=. Virorum celeberrimorum G. G. Leibnitii et Johannis Bernoullii Commercium philosophicum et mathematicum , volume=. 1745 , url=
-
[289]
arXiv preprint arXiv:2412.06264 , year=
Flow matching guide and code , author=. arXiv preprint arXiv:2412.06264 , year=
-
[290]
ICLR , year=
Simplifying, stabilizing and scaling continuous-time consistency models , author=. ICLR , year=
-
[291]
Medical image computing and computer-assisted intervention (MICCAI) , year=
U-net: Convolutional networks for biomedical image segmentation , author=. Medical image computing and computer-assisted intervention (MICCAI) , year=
-
[292]
ACM Transactions on Graphics (SIGGRAPH) , year=
Stylegan-xl: Scaling stylegan to large diverse datasets , author=. ACM Transactions on Graphics (SIGGRAPH) , year=
-
[293]
arXiv preprint arXiv:2407.02398 , year=
Consistency flow matching: Defining straight flows with velocity consistency , author=. arXiv preprint arXiv:2407.02398 , year=
-
[294]
arXiv preprint arXiv:2505.20297 , year=
DiSA: Diffusion Step Annealing in Autoregressive Image Generation , author=. arXiv preprint arXiv:2505.20297 , year=
Reviewed August 16, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.