DMA$^2$: Pixel-space Distribution Matching with Adversarial and Anchor Losses

Xin Lin; Zhifei Zhang; Yuqian Zhou; Haitian Zheng; Shaoteng Liu; Lehan Yang; Zhe Lin; Ming-Hsuan Yang

cs.CVSep 29, 2026

DMA $^2$ : Pixel-space Distribution Matching with Adversarial and Anchor Losses

Authors: Xin Lin, Zhifei Zhang, Yuqian Zhou, Haitian Zheng, Shaoteng Liu, Lehan Yang, Zhe Lin, Ming-Hsuan Yang, +1 more

Organizations: UC San Diego · Adobe Research · University of Virginia · UC Merced

Abstract

Distribution matching distillation (DMD) provides a general framework for few-step diffusion generation, but its modern text-to-image instantiations have been developed primarily around latent diffusion. It therefore overlooks key properties and design opportunities of native RGB. We revisit two DMD interfaces for pixel-space teachers. On the teacher-matching side, diagnostics show low-noise RGB matching is dominated by a local-texture cue, motivating a fixed high-noise matching band. On the real-data side, native clean-RGB outputs allow guidance from an external visual representation without traversing a decoder or sharing the heavy fake-score critic. DINO-Adv removes this critic from the adversarial gradient path and supplies local parametric patch guidance. For distribution-level guidance, we introduce AF-Loss, a parameter-free auxiliary semantic distribution-field objective designed for text-to-image DMD. It operates on detached rolling real and generated supports in the shared DINOv2 space while preserving prompt-conditioned teacher supervision. AF-Loss adds no learnable parameters or inference-time computation. Together these designs form DMA $^2$ . Across DPG-Bench, GenEval, VQAScore, and COCO30K, the four-step DMA $^2$ student performs better than the 25-step teacher and evaluated few-step distillers.

Figures & tables

Figure 1 : Qualitative results of four-step text-to-image generation of DMA 2 ( $512\times 512$ ).

Figure 2 : Overview of the proposed DMA 2 framework. DM-Band restricts teacher matching to a diagnostically selected high-noise range. For real-data guidance, DINO-Adv and AF-Loss share a frozen DINOv2 encoder and provide local patch guidance and a parameter-free semantic distribution-field target, respectively. Only $G_{\theta}$ , the fake-score critic, and the adversarial heads are trained; all auxiliary components are removed at inference.

Figure 3 : Separability diagnostic (C2ST). Each panel plots a held-out classifier’s real/fake separability (

y

-axis: ROC-AUC over

1{,}536

matched real/generated pairs,

0.5

is chance) against the DeCo matching level

t

(

x

-axis; larger

t

is less noise). The panels vary the setup: (a) the RGB input transform (raw texture, high-pass, low-pass,

32{\times}32

patch shuffle); (b) the generation stage

t_{g}\!\in\!\{0,.25,.5,.75\}

; (c) the input representation (RGB, the SDXL latent, frozen DINOv2). RGB separability changes sharply with

t

and is high only near clean, whereas the SDXL latent stays flat and DINOv2 stays separable into the high-noise band.

Figure 4 : Eq. 1 terms across $t_{m}$ ; $\Delta$ exhibits broader coherent patterns at high noise and sparse fine-scale patterns near clean. Each panel is independently normalized for spatial visualization; colors are not comparable in magnitude across cells.

DPG-Bench $\uparrow$	$U(.02,.98)$	$U(.02,.35)$
Pixel (DeCo)	82.55	82.96
Latent (SDXL)	67.64	63.23

Table 5

Adversary	$1$	$4$	$1$	$4$	s/step	#learn. [-1pt] (adv. path)
	DPG-Bench $\uparrow$		GenEval $\uparrow$		Train cost $\downarrow$
DMD2 Feat. GAN	78.35	82.12	0.6960	0.7709	1.82	560M
Frozen-DINOv2	81.26	82.55	0.7723	0.8155	0.72	1.58M

Table 6

Figure 5 : Anchor-Field Loss (AF-Loss). The generated image’s normalized final-layer DINOv2 CLS feature

z=\phi(\hat{x})

is attracted toward a rolling bank of real features

B_{r}

and repelled from a bank of generated features

B_{f}

, giving a displacement

V(z)

. AF-Loss converts this displacement into a stopped target whose gradient flows only through

z\!\to\!\hat{x}\!\to\!G_{\theta}

; the field itself has no learnable parameters. The banks are detached FIFO queues capped at

256

features.

Teacher DMD2 Decoupled DMA 2 (ours)	Teacher DMD2 Decoupled DMA 2 (ours)
“A set of four green plastic food containers displayed against a stark white background, each captured from a distinct angle to showcase the varying perspectives …”	“An up-close image showcasing the intricate interior of a walnut, split cleanly down the middle to reveal its textured, brain-like halves …”

“Three sleek, dark wooden boats are resting along the banks of a tranquil, azure blue lake, their oars tucked neatly inside …”	“A sprawling field blanketed with vibrant wildflowers, where a tall giraffe and a striped zebra stand side by side under acacia trees …”

Teacher DMD2 Decoupled DMA 2 (ours)	Teacher DMD2 Decoupled DMA 2 (ours)
“A cat on a leather chair next to remotes.”	“A large dog sitting on top of a roof.”

Figure 6 : Qualitative comparison at four steps. Top block: long, densely specified DPG-Bench prompts (abbreviated with “…”); bottom block: short COCO captions. DMA 2 preserves object count and composition and recovers fine structure, across both long and short text.

Model	NFE	GenEval $\uparrow$	DPG $\uparrow$	VQA $\uparrow$	IS $\uparrow$	CLIP $\uparrow$	Rec. $\uparrow$	NIQE $\downarrow$
DeCo ( Ma et al., 2026 ) (official)	25	0.8216	81.580	0.7014	35.58	0.3200	0.3110	4.064
Flash ( Chadebec et al., 2025 )	4	0.7814	76.703	0.6900	29.94	0.3178	0.1486	5.688
DMD2 ( Yin et al., 2024a )	4	0.7709	82.115	0.7058	36.76	0.3170	0.4349	4.168
TDM ( Luo et al., 2025 )	4	0.7924	82.367	0.7071	36.48	0.3165	0.4447	3.871
Decoupled ( Liu et al., 2026b )	4	0.8100	83.095	0.7111	36.51	0.3195	0.3516	3.953
DMDR ( Jiang et al., 2026 )	4	0.8063	83.160	0.7023	34.22	0.3198	0.4414	3.710

Table 1 : Main comparison. All few-step methods are re-implemented on the same teacher; bold/underline mark the best/second-best students at each step count.

Module			Metric
DM-Band	DINO-Adv	AF-Loss	GenEval $\uparrow$	DPG $\uparrow$	VQA $\uparrow$	IS $\uparrow$	Rec. $\uparrow$	NIQE $\downarrow$
			0.8037	81.115	0.6930	34.24	0.2802	5.174
	✓		0.8155	82.550	0.7045	39.38	0.4297	3.337
✓	✓		0.8193	82.959	0.7077	40.21	0.4318	3.034
		✓	0.8090	83.194	0.7079	39.46	0.3161	3.574
	✓	✓	0.8178	83.427	0.7113	39.52	0.4324	3.307

Table 2 : Component ablation. Bold/underline mark the best/second-best results.

Figure 7 : Hyperparameter sensitivity in DPG, GenEval, and IS; green bands mark defaults. These are OFAT sweeps around an anchor that is not the full model, so their default and zero-weight points are not directly comparable to the Table 2 component ablation (anchor protocol in Appendix D ).

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Method	Time (s/step) $\downarrow$
DMD2 ( Yin et al., 2024a )	1.82
Decoupled DMD ( Liu et al., 2026b )	1.82
DMDR ( Jiang et al., 2026 )	0.90
DMA 2 (ours)	0.72

Appendix

Table 3 : Same-pipeline training speed. All methods use the same hardware, global batch size, and end-to-end step-timing protocol. Lower is better.

Model	NFE	GenEval $\uparrow$	DPG $\uparrow$	VQA $\uparrow$	IS $\uparrow$	CLIP $\uparrow$	Rec. $\uparrow$	NIQE $\downarrow$
PixelGen-XXL teacher	25	0.7965	79.360	0.6804	37.24	0.3141	0.3328	4.064
DMA 2 -PixelGen (ours)	4	0.8112	81.708	0.6897	40.46	0.3184	0.4892	3.088
DMA 2 -PixelGen (ours)	1	0.8072	80.898	0.6961	39.48	0.3232	0.3872	3.275

Appendix

Table 4 : Extension to a PixelGen-XXL teacher: DMA 2 few-step students vs. the original PixelGen teacher. Metrics as in Table 1 ;

\uparrow

\downarrow

give direction, bold is best and underline second best per column. The four-step student uses the strictly-rerun schedule-matched replica.

DMD matching range	$U(.02,.20)$	$U(.02,.35)$	$U(.02,.70)$	$U(.15,.70)$	$U(.60,.90)$	$U(.02,.98)$	$U(.02,\{.25,.50,.75,.98\})$
	narrow	best	one-sided	wide	low-noise	full range	per-stage
GenEval $\uparrow$	0.8095	0.8193	0.8113	0.7924	0.7341	0.8155	0.8164
DPG-Bench $\uparrow$	82.720	82.959	82.426	82.038	80.330	82.550	82.018
CLIP $\uparrow$	0.3201	0.3225	0.3184	0.3213	0.3194	0.3214	0.3211
IS $\uparrow$	38.79	40.21	39.78	39.30	36.21	39.38	38.71
Recall $\uparrow$	0.4182	0.4318	0.4198	0.4062	0.3773	0.4297	0.4027

Appendix

Table 5 : Noise selection: matching-band position, width, and per-stage nesting. Four-step DINO-Adv students (

20

\lambda_{\mathrm{GAN}}{=}.01

, AF-Loss disabled) differ only in the DMD matching range;

\uparrow

\downarrow

give direction and bold is best per row. The high-noise band

U(.02,.35)

leads on nearly every alignment metric.

Objective	GenEval $\uparrow$	DPG $\uparrow$	VQA $\uparrow$	IS $\uparrow$	CLIP $\uparrow$	Rec. $\uparrow$	NIQE $\downarrow$
AF-Loss (CLS field, ours)	0.8090	83.194	0.7079	39.46	0.3246	0.3161	3.574
Direct DINOv2-CLS perceptual	0.8156	82.077	0.7028	37.85	0.3217	0.3146	4.675
DINOv2-CLS MMD (distribution matching)	0.8077	81.877	0.6949	35.77	0.3206	0.3330	4.433

Appendix

Table 6 : AF-Loss versus two alternative feature-space objectives on the same normalized last-layer CLS features, encoder, four-step protocol, and loss weight (

.05

): direct DINOv2-CLS perceptual matching, and a matched multi-scale DINOv2-CLS MMD (distribution matching) against the same

256

-feature real bank. Only the feature-space objective changes; bold is best per column.

AF feature	DPG $\uparrow$
DINOv2 $\{2,5,8,11\}$ spatial patch features (per-layer mean-pool, concat)	81.970
DINOv2 $\{2,5\}$ spatial patch-feature mean+std	81.430
DINOv2 $\{2,5\}$ spatial patch-feature quantile	81.760
DINOv2 normalized final-layer CLS feature (ours)	83.194

Appendix

Table 7: Anchor-Field feature ablation. All variants share the same field form at

\lambda_{\mathrm{AF}}\!=\!.05

; the normalized final-layer CLS feature is our default.

Anchor-Field term	DPG $\uparrow$
Attract only (real bank $B_{r}$ )	82.910
Repel only (fake bank $B_{f}$ )	80.170
Attract $+$ repel (ours)	83.194

Appendix

Table 8: Anchor-Field attract/repel ablation, at the default feature and

\lambda_{\mathrm{AF}}\!=\!.05

Teacher DMD2 Decoupled DMA 2 (ours)	Teacher DMD2 Decoupled DMA 2 (ours)
“An intricate oil painting that captures two rabbits standing upright in a pose reminiscent of the iconic American Gothic portrait, in early 20th-century rural clothing …”	“A rider atop a chestnut horse in the middle of a spacious pasture enclosed by a wooden fence, dotted with patches of green grass …”

“A frisky golden retriever with a shiny, shaggy coat stands next to a life-sized penguin statue in the midst of a bustling public park …”	“A vibrant yellow rabbit, its fur almost glowing with cheerfulness, bounds energetically across a sprawling meadow, its sizeable red-framed glasses slipping comically …”

Teacher DMD2 Decoupled DMA 2 (ours)	Teacher DMD2 Decoupled DMA 2 (ours)
“A polar bear walking over rocks in its enclosure.”	“A teddy bear wearing a green robe.”

Appendix

Figure 8 : Additional four-step qualitative comparisons under identical prompts (columns: 25-step teacher, DMD2, Decoupled, and DMA 2 ). Top block: long DPG-Bench prompts (abbreviated with “…”); bottom block: short COCO captions.

Explore similar work

Continuous-Time Distribution Matching for Few-Step Diffusion Distillation
May 7, 2026Tao Liu, Hao Yan, Mengting Chen +8Discrete Diffusion Model Interaction Trajectories
Improved Distributional Diffusion Models
Sep 29, 2026Tommaso Martorella, Alexandre Galashov, Felix Krause +4
Reinforcing Few-step Generators via Reward-Tilted Distribution Matching
May 25, 2026Yushi Huang, Xiangxin Zhou, Ruoyu Wang +3Text-To-Image Generative Models

May 7, 2026, cs.CVJ/K move · Enter open · S save

Continuous-Time Distribution Matching for Few-Step Diffusion Distillation

Tao Liu, Hao Yan, Mengting Chen, Taihang Hu+7

1VCIP, College of Computer Science, Nankai University · 2Alibaba Group · College of Artificial Intelligence, Jilin University

Discrete Diffusion Model Interaction Trajectories

Open paper AskPDF arXiv

Step distillation has become a leading technique for accelerating diffusion models, among which Distribution Matching Distillation (DMD) and Consistency Distillation are two representative paradigms. While consistency methods enforce self-consistency along the full PF-ODE trajectory to steer it toward the clean data manifold, vanilla DMD relies on sparse supervision at a few predefined discrete timesteps. This restricted discrete-time formulation and mode-seeking nature of the reverse KL divergence tends to exhibit visual artifacts and over-smoothed outputs, often necessitating complex auxiliary modules -- such as GANs or reward models -- to restore visual fidelity. In this work, we introduce Continuous-Time Distribution Matching (CDM), migrating the DMD framework from discrete anchoring to continuous optimization for the first time. CDM achieves this through two continuous-time designs. First, we replace the fixed discrete schedule with a dynamic continuous schedule of random length, so that distribution matching is enforced at arbitrary points along sampling trajectories rather than only at a few fixed anchors. Second, we propose a continuous-time alignment objective that performs active off-trajectory matching on latents extrapolated via the student's velocity field, improving generalization and preserving fine visual details. Extensive experiments on different architectures, including SD3-Medium and Longcat-Image, demonstrate that CDM provides highly competitive visual fidelity for few-step image generation without relying on complex auxiliary objectives. Code is available at https://github.com/byliutao/cdm.

Abstract

Explore similar work

Continuous-Time Distribution Matching for Few-Step Diffusion Distillation

Improved Distributional Diffusion Models

Reinforcing Few-step Generators via Reward-Tilted Distribution Matching