- Rating
- 3(+1)
- Confidence
- 2(−1)
- Informativeness
- 2
- Presentation
- 2(+1)
- Soundness
- 2
Read review
Paper summary
This paper proposes a diffusion distillation method StepX-OPD, which can distill multiple specialized teacher diffusion models into a 4-step student model in one stage. StepX-OPD combines several techniques from step distillation and on-policy distillation methods for diffusion models. The authors conduct experiments on two text-to-image generation benchmarks and compare the proposed method with sequential (two-stage) baselines. The paper also investigates the behavior of unintended rendering of propmpt words and explores a teacher-rollout target to mitigate this issue.
Critical strengths
- StepX-OPD combines capability consolidation and few-step distillation for diffusion models in one stage, which reduces the training costs and improves inference efficiency. The motivation is clear and the proposed method addresses a meaningful problem and could be practically useful for deployment.
- Text-to-image generation results on SD3.5-Medium and FLUX.1-dev demonstrate that StepX-OPD achieves comparable performance to specialized teacher models and improved overall performance over sequential (two-stage) OPD+TDM baselines.
- The investigation of unintended text generation with benchmark stratification and additional quality metrics is informative and is a nice addition to the main results.
Critical weaknesses
- The contribution of the proposed method is not clearly positioned in the literature. StepX-OPD combines several techniques from existing step distillation and on-policy distillation methods for diffusion models (e.g., SFT-augmented initialization, specialist teachers, source-specific training data and so on). It is unclear whether all of these techniques are necessary and how each of them affect the final performance. More ablation study is needed to isolated the core contribution of this paper from closely related methods.
- The presentation of the paper needs improving. Section 4 (method section) should start with a big picture of how the training recipe looks like and then describe each component in the method (e.g., why each component is necessary and how it works). Currently, the description of the proposed method is hard to follow. For example, it starts by explaining the difference between the schedule of CDM and TDM, which is very confusing.
- It is unclear whether the performance improvement is significant. For several quality measures (especially PickScore), StepX-OPD underperforms the best baselines on both benchmarks. Also, StepX-OPD's improved average score on FLUX.1-dev is mainly driven by the DeQA score. There is also a concrete contradiction: Appendix E says increasing OCR source’s weight improves both configurations in the text, but Table 11 shows the merged configuration’s OCR and Average score both noticeably decrease with increased OCR sampling.
Questions for authors
- Could the authors provide a matched comparison against routed TDM/CDM using exactly the same configurations as StepX-OPD (same initialization, teachers, prompt sources and training budget and so on)?
- Could the authors conduct an ablation study with ?
- For the two-stage sequential baseline, what if you first distill specialized teachers into few-step models and then use OPD to consolidate them into one model?
- Why do you only consider a 4-step student model in the experiment? How do StepX-OPD and other baselines perform for 1-step and 8-step settings?
8 comments
W3: It is unclear whether the performance gain is significant. On several quality metrics (especially PickScore), StepX-OPD is worse than the best baseline on both architectures, and the Average gain on FLUX.1-dev mainly comes from DeQA. In addition, the text of App. E says that raising the OCR source weight improves both configurations, while Table 11 shows that both OCR and Average of the merged-initialization configuration drop clearly. The two contradict each other.
A: Average is an aggregate of the four metrics, and the revision reports the four metrics as separate measurement columns. The raw values of StepX-OPD and the two sequential baselines in Tab. 1 are as follows.
| Architecture | Method | GenEval | OCR | DeQA | PickScore | Average |
|---|---|---|---|---|---|---|
| SD3.5-Medium | Flow-OPD + TDM | 0.875 | 0.751 | 4.217 | 22.60 | 0.8345 |
| SD3.5-Medium | DiffusionOPD + TDM | 0.838 | 0.833 | 3.952 | 23.41 | 0.8403 |
| SD3.5-Medium | StepX-OPD | 0.944 | 0.899 | 4.170 | 22.98 | 0.8901 |
| FLUX.1-dev | Flow-OPD + TDM | 0.864 | 0.901 | 4.087 | 23.30 | 0.8695 |
| FLUX.1-dev | DiffusionOPD + TDM | 0.882 | 0.925 | 4.005 | 23.32 | 0.8761 |
| FLUX.1-dev | StepX-OPD | 0.890 | 0.929 | 4.233 | 23.28 | 0.8902 |
On SD3.5-Medium, the GenEval and OCR of StepX-OPD are higher than those of the two sequential baselines. Its PickScore is lower than that of DiffusionOPD[3] + TDM, and its DeQA is lower than that of Flow-OPD[4] + TDM. On FLUX.1-dev, the GenEval, OCR, and DeQA of StepX-OPD are higher than those of the two sequential baselines, and its PickScore is lower than both.
On the Average gain on FLUX.1-dev, the gap in DeQA is the largest. StepX-OPD is 4.233, and the two sequential baselines are 4.087 and 4.005. At the same time, GenEval (0.890 versus 0.864 and 0.882) and OCR (0.929 versus 0.901 and 0.925) are also higher than those of the two sequential baselines.
On the contradiction in App. E, the reviewer's observation is correct. The text said that raising the OCR source weight improves both configurations, and the revision has corrected it so that it agrees with the table (Tab. 10 in the revision). In the merged-initialization configuration, the Average is 0.8901 at the source ratio 1:1:1 and 0.8427 at 1:2:1.
[3] DiffusionOPD: A Unified Perspective of On-Policy Distillation in Diffusion Models. SIGGRAPH Asia [4] Flow-OPD: On-Policy Distillation for Flow Matching Models. arXiv
Q1: Can the authors provide a comparison with routed TDM/CDM under exactly the same configuration (the same initialization, teachers, prompt sources, training budget, and so on)?
A: The cumulative ablation in W1 of this section already includes TDM + source routing under the same configuration. The table below adds CDM + source routing. The three rows have the same initialization, teachers, source ratio, and evaluation protocol.
| Configuration | GenEval | OCR | DeQA | PickScore | Average |
|---|---|---|---|---|---|
| TDM + source routing | 0.9263 | 0.8488 | 4.021 | 21.09 | 0.8476 |
| CDM + source routing | 0.9463 | 0.8288 | 4.021 | 22.49 | 0.8611 |
| StepX-OPD | 0.9442 | 0.8986 | 4.170 | 22.98 | 0.8901 |
The Average of routed TDM and routed CDM is 0.8476 and 0.8611, and that of StepX-OPD is 0.8901. All four metrics of routed TDM are lower than those of StepX-OPD. For routed CDM, only GenEval (0.9463) is slightly higher than that of StepX-OPD (0.9442), and the other three metrics are lower than those of StepX-OPD.
Q2: Can the authors give an ablation without velocity regression (the equation is not displayed in the original text, and from the context it is )?
A: Tab. 6 already gives this result, and the cumulative table in W1 of this section also contains this row. The three rows below change only velocity regression, and all other settings are the same.
| Velocity regression | GenEval | OCR | DeQA | PickScore | Average |
|---|---|---|---|---|---|
| Not used | 0.9434 | 0.8656 | 4.178 | 21.95 | 0.8722 |
| From step 0 | 0.9444 | 0.8950 | 4.177 | 22.96 | 0.8895 |
| From step 500 (StepX-OPD) | 0.9442 | 0.8986 | 4.170 | 22.98 | 0.8901 |
We thank the reviewer for reading our manuscript carefully and for these constructive comments. We respond to each point below.
W1: The position of the contribution in the literature is unclear. StepX-OPD combines several techniques from existing step distillation and on-policy distillation (SFT-augmented initialization, specialist teachers, source-based training data, and so on). It is unclear whether every component is necessary and how each one affects the final performance. More ablations are needed to separate the core contribution of this paper from related methods.
A: The revision states the origin of each component at the start of Sec. 4 and in Sec. 4.2. The distribution matching loss comes from TDM[1], the randomized grid is inspired by CDM[2], and the specialist teachers come from multi-teacher on-policy distillation. The core of this paper is that, within a single stage, source-routed specialists directly supervise the states of the deployed few-step student on its own rollout, and that a low-noise velocity regression motivated by a reverse KL decomposition is added. The table below adds the components one at a time. The three rows change only the listed component, and all other settings are the same.
After the randomized grid and interpolated re-noising are added, all four metrics rise, and Average rises from 0.8476 to 0.8722. After low-noise velocity regression is added, Average rises to 0.8901.
The other components also have corresponding ablations in the paper, summarized below.
[1] Learning Few-Step Diffusion Models by Trajectory Distribution Matching. ICCV [2] Continuous-Time Distribution Matching for Few-Step Diffusion Distillation. arXiv
W2: The presentation needs improvement. Sec. 4 should first give the overall picture of the training recipe, and then explain one by one why each component is necessary and how it works. It now opens with the difference between the CDM and TDM schedules, which is confusing.
A: The revision has reorganized Sec. 4 accordingly. A new opening paragraph outlines one training step. The training source selects a specialist, the specialist supervises states built on the student's own few-step rollout, and it acts through both the TDM distribution matching loss and low-noise velocity regression. The three components are then introduced in turn. The subsection order is now 4.1 source routing, 4.2 supervision-point construction, and 4.3 low-noise velocity regression. Sec. 4.2 now opens by stating directly on which grid the student rolls out and why states between grid nodes need to be built. The earlier comparison of the CDM and TDM schedules has been removed, and CDM is cited only as the origin of the randomized grid. The CFG target is now given in the paragraph and is no longer listed as a separate equation.