- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper studies temporal quantization of diffusion models, where the precision, the rounding rule (round-to-nearest or stochastic rounding), and an optional compensator can change across timesteps. The authors split the one-step quantization residual into a conditional bias and a zero-mean fluctuation, and derive a closed-form, high-probability bound on the terminal error under local Lipschitz and sub-Gaussian assumptions (Theorem 4.2). They then select the schedule that minimizes this bound under a computation budget, using dynamic programming with pruning. The parameters of the bound are estimated from samples and verified by split-conformal calibration. On CIFAR-10, SDXL-Turbo, SDXL, and PixArt-, the authors report that the bound covers the empirical errors and that the selected schedules reduce LPIPS and the terminal L2 error to the full-precision outputs.
Critical strengths
-
The method is supported by a theoretical analysis.
-
The method outperforms existing methods in the most settings.
Critical weaknesses
- The certification does not seem to follow directly from Theorem 4.2. Theorem 4.2 holds only if its parameters are upper bounds of their true values. In the paper, however, they are set to empirical statistics, which does not appear to guarantee this. For example:
-
: Assumption 4.1 requires the bias to be bounded by almost surely, while is set to the 0.8 quantile over probe states (Appendix E).
-
: The proof requires that each step amplifies the error by at most in every generation at deployment. , however, is set to 1.1 times the largest amplification observed during calibration under random schedules (Eq. (28)), which may be exceeded on new prompts or under the selected schedule. The factor of 1.1 also seems heuristic, and the paper does not explain how it is chosen or whether it is sufficient.
This affects the main claim of the paper, which is the certified guarantee. It would be helpful if the authors could clarify how the estimated parameters satisfy the assumptions of Theorem 4.2, and adjust the use of "certified" accordingly.
- The comparison may not fully account for cost.
-
The cost of each method is not reported. Tables 1 and 4 report quality metrics only, without the latency or memory of each method, and without the schedule and budget that TQDiff selects. TQDiff can use FP16 at some steps (Appendix F.2) and tends to select higher precision at early steps (Sec. 5.2), while most baselines use a single low precision. Fig. 4 compares the runtime with FP16 only, and the memory of TQDiff is higher than that of the single-precision baselines (Ln453-455). Part of the gain may therefore come from additional computation rather than from better scheduling. A similar question applies to TMPDiff [A], whose budget is not reported.
-
Some closely related methods are not compared. PTQD [B] also allocates bit widths across timesteps, and TAC-Diffusion [C] and AccuQuant [D] address the accumulation of quantization errors across steps. These methods are discussed in Secs. 1 and 2, but none of them seems to be evaluated.
Without this information, it is difficult to assess the trade-off between error and cost, which determines the value of a scheduling policy. It would be helpful to report the cost of each method in Tables 1 and 4, compare the methods under matched budgets, and include these methods if feasible.
References
[A] TMPDiff: Temporal Mixed-Precision for Diffusion Models, arXiv 2026
[B] PTQD: Accurate Post-Training Quantization for Diffusion Models, NeurIPS 2023
[C] Timestep-Aware Correction for Quantized Diffusion Models, ECCV 2024
[D] AccuQuant: Simulating Multiple Denoising Steps for Quantizing Diffusion Models, NeurIPS 2025
Questions for authors
Among DiTs, the experiments seem to be conducted on PixArt- only. Could the authors evaluate TQDiff on recent flow-matching DiTs such as FLUX.1 or SD3.5?
3 comments
Response to Reviewer 7MAr (Part 2)
- SDXL (, budget matched to TMPDiff): TQDiff selects 4 FP16 steps (0–3), 18 W8A8-RN steps (4–21), and 28 W4A4-RN steps (22–49), using fewer FP16 steps (4 vs. 12) and fewer average bits (6.4 vs. 6.9) than TMPDiff (12 FP16 + 38 W4A4-RN). Both schedules run in (vs. for FP16, for uniform W8A8, and for uniform W4A4), while TQDiff lowers LPIPS from 0.158 to 0.105 on 200 fresh prompts (paired ).
- PixArt- (, budget matched to TMPDiff): TQDiff uses 0 FP16 steps, selecting 10 W8A8-RN steps (3–12) and 10 W4A4-RN steps (0–2 and 13–19) (6.0 avg. bits vs. TMPDiff's 7.0 avg. bits with 5 FP16 steps at 2–6), reducing LPIPS from 0.227 to 0.183 () on 500 held-out prompts at vs. denoising latency.
- Six matched budgets on SDXL (Figure 4, right, p. 10; Table 8, Appendix I, p. 29): Across six budgets matching TMPDiff-family schedules with FP16 steps (planned cost differences ) plus uniform W8A8's budget (), TQDiff achieves paired LPIPS reductions of 0.028 to 0.069 at every budget, with all 95% bootstrap CIs excluding zero.
- Memory disclosure (Table 7, p. 28): All-resident SDXL TQDiff uses peak memory vs. for TMPDiff and for FP16 (on PixArt-, TQDiff uses vs. for TMPDiff and for FP16). Sequential residency lowers SDXL TQDiff to ( on PixArt-) at ( transfer overhead at batch size 1, or at batch size 8).
W2b: Comparisons with PTQD, TAC-Diffusion, and AccuQuant
Manuscript location: Revised in Section 2.2 & 2.3 (pp. 3–4), Section 5.2 (p. 10), and New Appendix K (Table 9, p. 30).
- Controlled comparison with PTQD [B] (Appendix K, Table 9, p. 30): We implemented PTQD's SNR-based mixed-precision allocation and affine step correction on our shared SDXL engine (using W8A8 in place of W4A8 for kernel compatibility, fitting on 64 prompts, and omitting noise-variance correction since DDIM has ). Under the exact same two-precision library and planned budget (), TQDiff achieves LPIPS 0.130 () vs. 0.369 () for PTQD's SNR ordering without correction and 0.371 () with correction (paired allocation improvement on 200 prompts; full-library TQDiff achieves 0.106). The schedule structure explains why: PTQD's local SNR criterion assigns W8A8 to the final 27 steps () where clean-signal SNR is highest, whereas TQDiff accounts for multi-step error amplification and assigns W8A8 to the initial 27 steps (), preventing early quantization errors from compounding.
- Relationship to TAC-Diffusion [C] and AccuQuant [D] (Section 2.2 & 2.3, pp. 3–4; Appendix K, p. 30): TAC-Diffusion (timestep-aware dynamic correction of quantization residuals and exposure bias) and AccuQuant (multi-step quantizer calibration) improve individual quantized operators rather than selecting temporal precision schedules. They are therefore orthogonal to TQDiff and can enter as candidate actions, just as TCEC does. We now explicitly introduce TAC-Diffusion in Section 2.2 (p. 3) and clarify its complementary role alongside AccuQuant in Section 2.3 (p. 4) without claiming empirical superiority over unevaluated operator-level methods.
Question: Applicability to Recent Flow-Matching DiTs (FLUX.1 or SD3.5)
Manuscript location: Revised in Section 2.1 (p. 3) and Section 6 Conclusion (p. 10).
Mathematically, deterministic flow-matching Euler updates take the form , which is a direct instance of the discrete-time update in Eq. (1). Thus, Theorem 4.2, Proposition 4.3, and our DP planner apply directly to flow-matching DiTs once per-step action profiles are collected. While our PixArt- experiments validate TQDiff on a Diffusion Transformer with a multistep solver, we have not yet conducted empirical evaluations on 12B-scale flow-matching models such as FLUX.1 or SD3.5, and we state this scope explicitly in Section 6 (p. 10).
I thank the authors for the detailed responses and additional experiments. However, in the revision, the certified guarantee comes from the split-conformal calibration newly introduced in Sec. 4.2, not from Theorem 4.2. Theorem 4.2 now serves mainly as a planning objective. Since the certified bound derived from the analysis is the main contribution of the paper, this change weakens the contribution. I therefore keep my rating.
Response to Reviewer 7MAr (Part 1)
We thank you for your constructive review and for highlighting the need to clarify how our theoretical bound relates to empirically estimated parameters and to evaluate TQDiff against related methods under matched computational budgets. All corresponding revisions are marked in blue in the manuscript.
W1: How Estimated Parameters Connect to Theorem 4.2 and Deployment Certification
You are completely right that Theorem 4.2 requires valid per-step upper bounds (ρk,βk,Σk), and that empirical 0.8-quantile biases and 1.1× secant heuristics alone do not establish almost-sure worst-case conditions. To address your request ("clarify how the estimated parameters satisfy the assumptions of Theorem 4.2, and adjust the use of 'certified' accordingly"), we formalize a two-stage certification framework and clarify the exact scope of "certified":
W2a: Reporting Costs, Schedules, and Matched-Budget Comparisons
To ensure that quality gains come from scheduling rather than extra compute or extra FP16 steps, Appendix H (Tables 6–7, pp. 27–28) reports the exact action libraries, budgets, selected schedules, FP16 step counts, average bit widths, measured denoising latencies, and peak GPU memory for all methods on a shared RTX 5070 Ti execution engine:
(the rest of the response is in the next block due to the characters limit.)