- Rating
- 4(+1)
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 3
Read review
Paper summary
This paper proposes SpurFlow, a few-step image editing distillation framework that addresses the imbalance between preserved and edited regions. It introduces learnable token-wise reference coefficients and residual velocity prediction to adaptively preserve source content, along with localized adversarial supervision to enhance editing quality.
Critical strengths
- The paper identifies the gradient imbalance between preserved and edited regions as an important limitation of conventional distillation methods for image editing.
- The spatially adaptive reference coefficients and localized adversarial supervision provide a practical approach to improving few-step image editing without explicit edit-mask annotations.
Critical weaknesses
- This work identifies gradient dominance by preserved regions as a key limitation of conventional distillation, but does not directly demonstrate how SpurFlow redistributes gradient contributions between preserved and edited regions. Quantitative gradient analysis after introducing the proposed reference coefficients would support the central motivation.
- Table 2 shows that learnable reference coefficients alone provide only modest improvements. The author should support effectiveness of the proposed velocity parameterization and the improvements stem from localized adversarial supervision rather than the core distillation mechanism.
Questions for authors
- Could the authors provide quantitative gradient analysis to demonstrate how SpurFlow redistributes gradient contributions between preserved and edited regions?
- Could the authors disentangle the contributions of the proposed velocity parameterization and localized adversarial supervision?
4 comments
Response to Reviewer 3Zie (2/2)
Component Contributions (Answer to W2 and Q2)
We thank the reviewer for raising this question and kindly invite the reviewer to revisit Sec. 5.2 and Table 2 for the existing component ablations. For clarity, we summarize them below, include the previously omitted DMD2 [1] baseline, and provide additional ablations on FLUX.2-Klein.
Experimental setup
We evaluate six configurations to distinguish the contributions of the velocity parameterization and localized adversarial supervision. The configuration IDs below are shared across both backbones. Random-Crop GAN and α-Guided GAN both retain global adversarial supervision.
| ID | Configuration |
|---|---|
| A | DMD2 (full-field velocity) |
| B | Reference formulation (fixed coefficients) |
| C | B + Learnable token-wise coefficients |
| D | C + Global GAN |
| E | C + Random-Crop GAN |
| Ours | C + α-Guided GAN (SpurFlow) |
In the following tables, Action through Average are ImgEdit-Bench scores; Q-SC, Q-PQ, and Q-O are GEdit-Bench v2 scores. Higher scores are better; lower NFE is better. Score pairs in the subsequent analysis denote ImgEdit-Bench average / GEdit-Bench Q-O.
Component analysis on FLUX.1-Kontext
| ID | NFE | Action | Add | Adjust | Background | Compose | Extract | Remove | Replace | Style | Average | Q-SC | Q-PQ | Q-O |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| A | 4 | 3.92 | 3.95 | 3.69 | 3.05 | 3.09 | 1.81 | 2.31 | 3.26 | 3.96 | 3.17 | 5.74 | 7.27 | 5.39 |
| B | 2 | 3.95 | 3.72 | 3.85 | 3.55 | 2.83 | 2.30 | 3.25 | 4.02 | 4.01 | 3.49 | 6.01 | 7.65 | 5.72 |
| C | 2 | 4.13 | 3.87 | 3.91 | 3.69 | 2.90 | 2.26 | 3.58 | 4.10 | 3.97 | 3.59 | 6.03 | 7.70 | 5.76 |
| D | 2 | 4.04 | 3.96 | 3.87 | 3.66 | 3.38 | 2.41 | 3.44 | 4.18 | 3.98 | 3.63 | 6.01 | 7.80 | 5.78 |
| E | 2 | 3.71 | 3.81 | 4.12 | 3.76 | 3.03 | 2.05 | 3.84 | 4.19 | 3.94 | 3.62 | 5.99 | 7.79 | 5.77 |
| Ours | 2 | 4.27 | 4.08 | 4.19 | 4.03 | 3.49 | 2.52 | 3.55 | 4.17 | 4.02 | 3.81 | 6.10 | 8.03 | 5.93 |
Component analysis on FLUX.2-Klein
| ID | NFE | Action | Add | Adjust | Background | Compose | Extract | Remove | Replace | Style | Average | Q-SC | Q-PQ | Q-O |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| A | 2 | 3.82 | 4.12 | 3.70 | 3.56 | 3.01 | 2.19 | 2.40 | 3.65 | 4.56 | 3.46 | 7.17 | 7.40 | 6.73 |
| B | 2 | 4.36 | 4.39 | 4.37 | 4.25 | 3.53 | 2.31 | 4.37 | 4.40 | 4.77 | 4.04 | 7.42 | 7.98 | 7.06 |
| C | 2 | 4.55 | 4.31 | 4.45 | 4.31 | 3.62 | 2.26 | 4.61 | 4.42 | 4.80 | 4.11 | 7.48 | 8.06 | 7.14 |
| D | 2 | 4.52 | 4.28 | 4.33 | 4.30 | 3.65 | 2.28 | 4.08 | 4.45 | 4.72 | 4.04 | 7.42 | 8.13 | 7.23 |
| E | 2 | 4.56 | 4.34 | 4.26 | 4.34 | 3.32 | 1.98 | 2.73 | 4.06 | 4.73 | 3.78 | 7.19 | 8.09 | 6.95 |
| Ours | 2 | 4.57 | 4.37 | 4.64 | 4.52 | 3.65 | 2.23 | 4.52 | 4.56 | 4.85 | 4.21 | 7.60 | 8.08 | 7.31 |
(1) Velocity parameterization.
Rows B and C evaluate the reference formulation without adversarial supervision. On FLUX.1-Kontext, introducing the reference term with fixed coefficients improves scores from 3.17/5.39 to 3.49/5.72, while reducing NFE from 4 to 2. Learning the token-wise coefficients further improves them to 3.59/5.76. On FLUX.2-Klein, the corresponding progression is 3.46/6.73 → 4.04/7.06 → 4.11/7.14. These results demonstrate improvements from the reference-based parameterization before localized adversarial supervision is introduced.
(2) Localized adversarial supervision.
Configurations C, D, E, and Ours retain the same reference-based parameterization and compare no GAN, Global GAN, Random-Crop GAN, and our coefficient-guided GAN at 2 NFE. Adding our localized GAN improves scores from 3.59/5.76 to 3.81/5.93 on FLUX.1-Kontext and from 4.11/7.14 to 4.21/7.31 on FLUX.2-Klein. It also outperforms both Global GAN and Random-Crop GAN on these two metrics across both backbones, supporting the benefit of directing adversarial supervision toward edited regions.
Overall, the reference-based velocity parameterization benefits both backbones, with larger gains on FLUX.2-Klein. Localized adversarial supervision provides further gains on both: its ImgEdit improvement is more pronounced on FLUX.1-Kontext (+0.22), while remaining beneficial on FLUX.2-Klein (+0.10), and both backbones gain 0.17 in Q-O. These results support the complementary contributions of the two designs.
We have revised the paper accordingly. The additional gradient analysis is included in Appendix G, “Regional Gradient Analysis,” and the additional FLUX.2-Klein ablations are included in the appendix section “Component Ablations on FLUX.2-Klein.” Changes in the revised manuscript are highlighted in blue.
[1] Yin et al. Improved Distribution Matching Distillation for Fast Image Synthesis. NeurIPS, 2024.
I thank the authors for the detailed clarifications and additional experiments. The Appendix G gradient analysis and the new component ablations do make the mechanism and the individual contributions clearer. I will raise my score to 4.
Dear Reviewer,
We thank the reviewer for taking the time to revisit our analysis and additional experiments. We appreciate the constructive feedback throughout the discussion, which has helped us clarify our method and strengthen the paper.
Thank you again for your time and consideration.
Best regards, Authors
Response to Reviewer 3Zie (1/2)
We sincerely thank Reviewer 3Zie for carefully reading our paper and providing constructive feedback. We address the reviewer's concerns and questions point by point below.
Gradient Redistribution (Answer to W1 and Q1)
We address this concern through the analytical derivation already provided in Appendix B and additional empirical gradient analysis.
(1) Analytical Gradient Derivation in Appendix B
We thank the reviewer for this suggestion and kindly invite the reviewer to revisit Appendix B, which provides an exact token-wise derivation of gradients with respect to the shared trainable parameters. Eqs. (18)–(22) decompose the gradient into reference-related and remaining-mismatch contributions, showing how the reference-related contribution is attenuated as the coefficients approach full reference reuse. Eqs. (23)–(24) further show that this contribution vanishes for ideally preserved tokens, while unresolved edits retain their residual target-error signal. This explains how our formulation can reduce preservation-related gradients and increase the relative influence of edited regions. We will highlight this analysis more clearly in the main text.
(2) Additional Empirical Gradient Analysis
To complement the derivation, we compare regional backbone gradients between SpurFlow and ArcFlow on FLUX.1-Kontext. SpurFlow uses a 20k-step warmup without GAN supervision, followed by the mask-guided GAN stage. We measure gradients at both checkpoints with respect to the shared backbone only, excluding the GAN loss and all discriminator parameters.
To ensure a consistent comparison, all three configurations use the same pool of 32k training pairs and the same token partition. We generate the edit masks once using SpurFlow and reuse them for ArcFlow and both SpurFlow checkpoints. Measurements cover 5000 gradient-measurement steps on 32 GPUs, with one pair per GPU per step. All regional losses use the same whole-image normalization.
Let ϕ denote the shared backbone parameters. For each region (r = pres or edit), we compute the average per-pair gradient norm:
Gr=N1∑n=1N∥∇ϕLn,r∥2.
Here, N is the number of pair-level gradient measurements, and the regional loss is computed for each measured training pair. Relative edit gradient magnitude (%) is defined as:
Redit=100×Gpres+GeditGedit.
The table below reports regional backbone gradient norms on the same training-pair pool. Masks are generated once by SpurFlow and reused across all rows. Reported gradients exclude the GAN loss and discriminator parameters. Each entry averages per-pair backbone gradient norms.
SpurFlow's relative edit gradient magnitude is 60.74% after warmup and increases to 67.89% at the GAN-stage checkpoint, compared with 52.45% for ArcFlow. These results indicate greater relative gradient emphasis on the predicted edit regions under the shared evaluation protocol.