- Rating
- 3(+1)
- Confidence
- 4
- Informativeness
- 1
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The paper studies offline MBO through ranking errors between near-optimal and inferior candidates. Its analysis separates candidate reachability from final selection and connects selection error to training-pair mismatch. It proposes DAR, which emphasizes comparisons between high-quality observations and the remaining data, with some comparisons within the high-quality subset. Experiments include Branin and five Design-Bench tasks.
Critical strengths
-
I find the focus on which comparisons matter useful. Separating reachability from selection makes the scope of the ranking argument clearer and provides a reasonable motivation for concentrating training on promising designs.
-
I like the simplicity of DAR. It uses existing labels and a straightforward sampling change, and achieves competitive results across continuous and discrete tasks with shared hyperparameters and ten-seed reporting.
Critical weaknesses
-
I am not convinced that the experiments isolate the proposed contribution beyond RaM. The comparison changes both the ranking loss and pair sampling, while the ablations never compare DAR against global sampling with the same hinge loss. The theory also does not establish that DAR reduces the target-distribution mismatch relative to global sampling. Moreover, best-of-128 ground-truth evaluation does not directly test the surrogate-based selection studied by the theory.
-
I find the baseline comparison incomplete, particularly for the closest prior work. Table 1 uses RaM-RankCosine but omits RaM-ListNet, whose published means exceed DAR on three of the five tasks. Recent methods such as SPADE [1] and established baselines such as COMs [2] and BDI [3] are also absent. Given the small improvements over the included RaM variant, I need stronger comparisons to judge the practical advance.
[1] https://arxiv.org/abs/2605.11246 [2] https://arxiv.org/abs/2107.06882 [3] https://arxiv.org/abs/2209.07507
Questions for authors
-
Could you compare global pairing and DAR pairing while keeping the hinge loss, initialization, training budget, and search settings identical? This would help isolate the benefit of the proposed sampling strategy.
-
Why is RaM-ListNet omitted? Does DAR retain an advantage against it and SPADE under matched evaluation settings? Please also identify which Table 1 results were reproduced versus inherited.
-
Could you report the true score of the surrogate-selected candidate, alongside the best-of-128 score? This would directly test whether the proposed training improves the selection behavior analyzed in the paper.
5 comments
Response by the authors (2/3)
W2 & Q2 Comparisons with RaM-ListNet, SPADE, COMs, BDI, and DynAMO, and result provenance
W2: I find the baseline comparison incomplete, particularly for the closest prior work. Table 1 uses RaM-RankCosine but omits RaM-ListNet, whose published means exceed DAR on three of the five tasks. Recent methods such as SPADE [1] and established baselines such as COMs [2] and BDI [3] are also absent. ...
Q2: Why is RaM-ListNet omitted? Does DAR retain an advantage against it and SPADE under matched evaluation settings? Please also identify which Table 1 results were reproduced versus inherited.
Thank you for pointing these out. Because DAR's training objective is fundamentally a pairwise ranking loss whereas ListNet is a listwise loss, our initial submission selected RaM-RankCosine for a fair comparison under pairwise supervision. Following your and Reviewer miZM's suggestion, we expand our evaluation to 29 methods across 12 tasks (5 Design-Bench, 4 GTOPX trajectory, and 3 ViennaRNA tasks), adding RaM-ListNet [4], BDI [3], COMs [2], DynAMO [5], and SPADE [1] under matched 128-candidate evaluation (when official repositories such as SPADE do not release per-task hyperparameter configurations, tuning is performed strictly on an 80/20 held-out split of each task's offline dataset without oracle queries). Across all 29 methods and 12 tasks (complete 29 × 12 per-task table in https://anonymous.4open.science/r/DAR-iclr27/table_r1_full_results.md and average-rank chart in https://anonymous.4open.science/r/DAR-iclr27/figures/fig_r2_12task_avg_rank.png), DAR+ ranks 1st overall (6.58) and DAR ranks 2nd, outperforming other methods while avoiding ListNet's listwise training overhead, as shown below.
| Method | Ant | D'Kitty | Supercon | TF-Bind-8 | TF-Bind-10 | 12-Task Avg. Rank (of 29) ↓ |
|---|---|---|---|---|---|---|
| DAR+ (Ours)* | 0.916±0.037 | 0.964±0.010 | 0.481±0.010 | 0.991±0.000 | 0.660±0.028 | 6.58 |
| DAR (Ours)* | 0.941±0.013 | 0.959±0.006 | 0.506±0.008 | 0.988±0.008 | 0.686±0.056 | 8.00 |
| ROOT* | 0.958±0.012 | 0.967±0.001 | 0.451±0.032 | 0.977±0.015 | 0.652±0.020 | 8.58 |
| RaM-RankCosine* | 0.915±0.026 | 0.949±0.014 | 0.522±0.020 | 0.976±0.023 | 0.663±0.047 | 8.83 |
| BDI* | 0.958±0.000 | 0.941±0.000 | 0.506±0.009 | 0.973±0.000 | 0.629±0.036 | 8.83 |
| RaM-ListNet* | 0.928±0.013 | 0.962±0.012 | 0.516±0.030 | 0.963±0.018 | 0.654±0.020 | 11.08 |
| DynAMO* | 0.874±0.022 | 0.940±0.004 | 0.517±0.021 | 0.882±0.093 | 0.557±0.032 | 13.83 |
| COMs* | 0.947±0.026 | 0.952±0.002 | 0.516±0.017 | 0.744±0.130 | 0.625±0.023 | 18.71 |
| SPADE* | 0.689±0.132 | 0.849±0.038 | 0.339±0.043 | 0.482±0.086 | 0.487±0.059 | 26.50 |
Regarding result provenance across the 29-method, 12-task benchmark (https://anonymous.4open.science/r/DAR-iclr27/table_r1_full_results.md), 25 of the 29 methods (including DAR+, DAR, ROOT, RaM-RankCosine, RaM-ListNet, BDI, DynAMO, COMs, SPADE, GA, GA-min, GA-mean, Match-OPT, ICT, DEMO, CMA-ES, CbAS, REINFORCE, BONET, RoMA, DDOM, MINs, BO-qEI, GABO, and FGM):
- discrete TF-Bind-8/10 for GTG, TF-Bind-10 for Tri-Mentoring, and all GTOPX and RNA-Binding tasks across all 29 methods are executed end-to-end by us (*)
- only the remaining Design-Bench entries marked with † (PGS and IOM on Design-Bench, GTG on Ant/D'Kitty/Superconductor, and Tri-Mentoring on Ant/D'Kitty/Superconductor/TF-Bind-8) are adopted directly from RaM [4];
- because training those remaining baselines across all tasks and seeds is computationally resource-intensive and time-consuming, we could not complete every seed within our limited compute budget during the rebuttal period; we ran 1–2 seeds and observed results closely matching RaM's reported scores, so we directly adopted RaM's Design-Bench numbers for those † entries for now, and we will continue the remaining runs and update them in a subsequent revision.
We have updated Section 5.2 and marked these result sources in the caption of Table 5 (Appendix I.2) in the revised paper. Thank you very much.
Response by the authors (3/3)
W1-3 & Q3 True score of the surrogate-selected candidate
W1 (excerpt): Moreover, best-of-128 ground-truth evaluation does not directly test the surrogate-based selection studied by the theory.
Q3: Could you report the true score of the surrogate-selected candidate, alongside the best-of-128 score? This would directly test whether the proposed training improves the selection behavior analyzed in the paper.
Thank you for this question. Alongside Best-of-128 (), we evaluate candidate-batch median quality and the true score of the single candidate chosen by the frozen surrogate (), including DAR+ with early-epoch synthetic calibration (applying the synthetic pair mixture during the first 50 training epochs and real DAR pairs during the remaining 50 epochs). Results across the five Design-Bench tasks are reported in the table below (and plotted in https://anonymous.4open.science/r/DAR-iclr27/figures/fig_r4_surrogate_selection.png).
| Evaluation Metric | Method | Ant | D'Kitty | Supercon | TF-Bind-8 | TF-Bind-10 | Avg. Rank ↓ |
|---|---|---|---|---|---|---|---|
| Median-of-128 | MSE | 0.559±0.020 | 0.882±0.002 | 0.357±0.012 | 0.510±0.023 | 0.468±0.008 | 2.60 |
| Median-of-128 | RaM-ListNet | 0.581±0.016 | 0.885±0.003 | 0.348±0.009 | 0.557±0.026 | 0.465±0.007 | 2.20 |
| Median-of-128 | DAR (Ours) | 0.562±0.021 | 0.887±0.002 | 0.374±0.016 | 0.585±0.028 | 0.472±0.006 | 1.20 |
| Surrogate Top-1 () | MSE | 0.586±0.237 | 0.873±0.024 | 0.352±0.039 | 0.817±0.150 | 0.469±0.077 | 2.40 |
| Surrogate Top-1 () | RaM-ListNet | 0.732±0.214 | 0.877±0.014 | 0.341±0.054 | 0.862±0.090 | 0.458±0.043 | 2.00 |
| Surrogate Top-1 () | DAR (Ours) | 0.596±0.223 | 0.881±0.032 | 0.367±0.033 | 0.723±0.189 | 0.489±0.050 | 1.60 |
| Surrogate Top-1 () | DAR+ (Ours) | 0.626±0.184 | 0.925±0.038 | 0.376±0.031 | 0.905±0.053 | 0.520±0.043 | 1.20 |
DAR achieves the best average rank on both candidate-batch median quality and single-candidate selection among the pure forward surrogates, and DAR+ further improves on every task. We have added these metrics alongside Best-of-128 in Section 5.3 and Appendix I.3. Thank you very much.
References
[1] Support-Proximity Augmented Diffusion Estimation for Offline Black-Box Optimization. https://arxiv.org/abs/2605.11246
[2] Conservative Objective Models for Effective Offline Model-Based Optimization. ICML 2021. https://arxiv.org/abs/2107.06882
[3] Bidirectional Learning for Offline Infinite-Width Model-Based Optimization. NeurIPS 2022. https://arxiv.org/abs/2209.07507
[4] Offline Model-Based Optimization by Learning to Rank. ICLR 2025.
[5] Michael S. Yao, James C. Gee, and Osbert Bastani. Diversity By Design: Leveraging Distribution Matching for Offline Model-Based Optimization. ICML 2025.
[6] ROOT: Rethinking Offline Optimization as Distributional Translation via Probabilistic Bridge. NeurIPS 2025.
We hope our response has addressed your concerns. If we have missed anything, please let us know.
Response by the authors (1/3)
Thank you for your comments. Below please find our responses.
W1-1 & Q1 Global pairing vs. DAR pairing under the same hinge loss
Thank you for suggesting this ablation. We run a controlled comparison keeping the pairwise hinge loss (β=0.4), model initialization, training pair budget, and search settings identical, changing only global uniform pairing vs. DAR pairing. Results are reported in the table below (and visualized in https://anonymous.4open.science/r/DAR-iclr27/figures/fig_r3_global_vs_dar.png).
Replacing global pairing with DAR reduces top-versus-rest inversion error (6.45% to 1.36% on average) and improves within-top Spearman rank correlation (0.429 to 0.689) on every task, improving average Best-of-128 score (0.798 to 0.811) under identical loss and budget. We have added this comparison to Section 5.3 and Appendix I.3, separate from the hyperparameter ablations. Thank you very much.
W1-2 Theoretical condition for reducing target-distribution mismatch
Thank you for pointing this out. For a fixed target distribution Qε with finite first moments under the same pair metric, let
e=W1(Qε,Qτ),d=W1(Qτ,Qglobal),a=λW1(Qτ,Qintra).
By the triangle inequality on the mixture QDAR=(1−λ)Qτ+λQintra, we have W1(Qε,QDAR)≤e+a, while the reverse triangle inequality gives W1(Qε,Qglobal)≥d−e. Thus, d>2e+a is sufficient for W1(Qε,QDAR)<W1(Qε,Qglobal): whenever the empirical cross-region proxy Qτ is closer to Qε than half its distance to global pairing (up to the λ=0.1 within-top term a), DAR strictly reduces Wasserstein mismatch. We have added this derivation to Appendix E. Thank you very much.