- Rating
- 3
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes SABR, a training-free sampling method for improving tail-class diversity in diffusion models. The authors identify stronger local conditional responses in tail classes and introduce pressure-guided block selection with SNR-dependent conditioning modulation. Experiments demonstrate improved tail-class generation quality and mode coverage across several benchmarks.
Critical strengths
- The analysis of frequency-dependent conditional pressure provides an interesting perspective on the concentration of tail-class patterns.
- SABR improves the generation of tail classes without retraining the diffusion model.
- The experiments cover multiple datasets and baseline methods.
Critical weaknesses
- Novelty in methodology is moderate. SABR mostly combines existing conditional adjustment ideas, while introducing heuristic pressure thresholds and a gate control mechanism based on signal-to-noise ratio.
- Component-level verification is insufficient. The independent contributions of signal-to-noise ratio adaptation, pressure budget, and routing decisions have not been clearly separated.
- Generalization evidence is limited. The main pattern coverage analysis focuses on CIFAR-100-LT, and its applicability to modern diffusion Transformer architectures remains unclear.
Questions for authors
Please see the weakness part.
2 comments
Weakness 3 Generalization across datasets and to DiT
Thank you for asking for more generalization evidence. We added tests of real-mode coverage on another dataset and generation performance with another architecture. The original real-mode analysis focused on CIFAR-100-LT; we now add tail-class mode coverage on CIFAR-10-LT. We also compare DiT with DiT (SABR) on both CIFAR-10-LT and CIFAR-100-LT at 100, 200, 300, 400, and 500 epochs. The results below address these two aspects in turn.
The CIFAR-10-LT analysis compares DDPM and DDPM + SABR using four real-mode metrics. Together with the original CIFAR-100-LT analysis, it tests whether the mode-coverage gains carry over to another dataset.
Table 2. Tail-class real-mode coverage on CIFAR-10-LT. Changes are SABR minus DDPM. Better results are bold.
| Metric | DDPM | DDPM+SABR | Change |
|---|---|---|---|
| Effective mode ratio ↑ | 0.7144 | 0.7367 | +0.0223 |
| Mode-mass recall ↑ | 0.6821 | 0.7062 | +0.0241 |
| JS divergence ↓ | 0.1151 | 0.0966 | -0.0185 |
| Maximum mode-share ratio ↓ | 2.3462 | 2.2043 | -0.1419 |
The effective mode ratio and mode-mass recall rise by 0.0223 and 0.0241, while JS divergence and the maximum mode-share ratio fall by 0.0185 and 0.1419. All four metrics change in the same direction as on CIFAR-100-LT in the paper. The DiT results below examine transfer across architectures.
Table 3. DiT and DiT (SABR) on CIFAR-10-LT at different training epochs. Better results are bold.
| Epoch | Method | Overall FID ↓ | Overall Recall ↑ | Few FID ↓ | Few Recall ↑ |
|---|---|---|---|---|---|
| 100 | DiT | 55.255 | 31.55% | 83.901 | 35.47% |
| 100 | DiT (SABR) | 55.113 | 31.79% | 83.844 | 35.75% |
| 200 | DiT | 30.503 | 34.66% | 37.541 | 36.52% |
| 200 | DiT (SABR) | 29.805 | 35.21% | 36.740 | 37.59% |
| 300 | DiT | 20.237 | 35.15% | 23.844 | 30.05% |
| 300 | DiT (SABR) | 19.131 | 36.82% | 23.632 | 31.75% |
| 400 | DiT | 17.332 | 34.82% | 24.863 | 25.39% |
| 400 | DiT (SABR) | 16.596 | 38.04% | 18.788 | 34.15% |
| 500 | DiT | 14.648 | 35.97% | 23.567 | 24.35% |
| 500 | DiT (SABR) | 14.115 | 39.43% | 17.918 | 33.02% |
Table 4. DiT and DiT (SABR) on CIFAR-100-LT at different training epochs. Better results are bold.
| Epoch | Method | Overall FID ↓ | Overall Recall ↑ | Few FID ↓ | Few Recall ↑ |
|---|---|---|---|---|---|
| 100 | DiT | 43.151 | 31.49% | 56.005 | 32.52% |
| 100 | DiT (SABR) | 43.270 | 31.48% | 56.208 | 32.53% |
| 200 | DiT | 25.704 | 36.31% | 39.758 | 37.76% |
| 200 | DiT (SABR) | 25.484 | 36.44% | 38.660 | 38.04% |
| 300 | DiT | 19.124 | 37.40% | 31.315 | 39.58% |
| 300 | DiT (SABR) | 18.899 | 37.35% | 31.103 | 39.54% |
| 400 | DiT | 16.318 | 36.73% | 27.781 | 36.35% |
| 400 | DiT (SABR) | 16.109 | 37.78% | 26.208 | 38.25% |
| 500 | DiT | 14.269 | 34.87% | 24.737 | 31.62% |
| 500 | DiT (SABR) | 13.605 | 37.04% | 22.279 | 36.84% |
SABR gives clear tail-class gains on both datasets. At 500 epochs, Few FID falls from 23.567 to 17.918 on CIFAR-10-LT, while Recall rises from 24.35% to 33.02%. On CIFAR-100-LT, Few FID falls from 24.737 to 22.279 and Recall rises from 31.62% to 36.84%. Overall FID and Recall also improve on both datasets. Results at the other epochs show gains at most training stages. These findings extend the results from the original U-Net models to the tested DiT models on two long-tailed datasets. We will add these experiments to the revised paper.
Thank you again for your time and constructive feedback.
Thank you for your careful review, your positive comments on our method, and your suggestions on novelty, component analysis, and generalization. We have added experiments and analysis in response. Below, we address the three main weaknesses, which also cover the items referred to in the questions-for-authors section.
Responses to the main weaknesses
Weakness 1 Method novelty and design rationale
SABR builds on the broader idea of adjusting conditioning. Its specific design follows a sequence of measurements and intervention tests aimed at understanding tail-class generation.
We first observed limited mode coverage and stronger concentration in tail-class outputs. To investigate this, we measured the direct class-conditional response at a fixed block input and found higher local pressure in tail classes, especially in Down-Late. Frequency reassignment then linked this higher pressure to training frequency, directing our attention to frequency-related local conditional responses. We next removed direct class conditioning in Down-Late. Tail-class denoising MSE rose by 18.31%, while FID and Recall improved and Precision fell. Reducing local conditioning can therefore improve coverage at a cost to fidelity. This led us to design a method that chooses where to intervene, limits the reduction, and controls its timing.
SABR uses pressure statistics from data-rich classes to build a frozen reference. Normalized block pressure determines intervention priority, a cumulative squared-pressure budget limits the adjustment, and the remaining excess determines the gate through an analytical rule. SNR gating controls when intervention is allowed. The budget comparison also gives lower Few FID and higher Recall with a cumulative budget than with a per-step budget, alongside slightly lower Precision.
The contribution is the finding that local conditional responses vary with training frequency and a method that turns this finding into reference-based control over intervention location, strength, and timing. Pressure measurement, frequency reassignment, and local intervention connect the observed tail-class behavior to the design of SABR.
Weakness 2 Contributions of the individual components
Thank you for suggesting a clearer component analysis. We added a 2 × 2 comparison of cumulative versus per-step budgets and γ = 0 versus γ = 1. We also tested block-position permutation and intervention in only the first block. These tests examine the roles of budget history, timing, and spatial routing. All settings use the same frozen CIFAR-100-LT DDPM, 50-step DDIM, CFG scale 2.5, paired initial noise, and 500 images per class.
Table 1. Component comparisons for pressure budgets, SNR gating, and block routing on CIFAR-100-LT. Better results are bold.
Note: Better results are bold.
The three components play different roles. For the pressure budget, cumulative control gives lower FID than per-step control at both γ settings, showing the value of retaining information from earlier interventions. For SNR gating, γ = 1 lowers FID under both budget rules. With the cumulative budget, Few Recall rises from 32.75% to 34.35%, showing that intervention timing affects the result. For block routing, permuting the block positions or using only the first block raises Few FID and lowers Recall. Thus, both the selected locations and the distribution of intervention across blocks matter.