Pith. sign in

REVIEW 4 major objections 5 minor 66 references

A formulation that prioritizes rare, promising samples reaches requirement-satisfying designs with fewer expensive simulations.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 08:05 UTC pith:UF3VOJNB

load-bearing objection A genuinely new anchor-state criterion for formulation search, but the efficiency claim rests on single runs and an in-sample selection that needs resampling tests. the 4 major comments →

arxiv 2607.21220 v1 pith:UF3VOJNB submitted 2026-07-23 cs.NE

Search Hardness-Aware LLM-Based Problem Formulation for Expensive Simulation-Driven Design

Yuchen Li , Handing Wang , Bing Xue , Mengjie Zhang This is my paper
classification cs.NE
keywords search-hardness-aware formulationLLM-based problem formulationexpensive simulation-driven designanchor state selectionsatisfaction statesevolutionary formulation searchantenna design optimizationsurrogate-assisted optimization
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper is trying to establish that the choice of problem formulation—not just the optimizer—controls how many expensive simulator calls are needed, and that a good formulation can be found automatically by reading search hardness out of the first batch of evaluated designs. It argues that initial space-filling evaluations contain partial paths to the target region: frequently observed satisfaction states are easy for an optimizer to reach anyway, whereas rare states mark hard but promising directions. The proposed method, SHA-PF, selects the rarest non-dominated state as an anchor and searches over LLM-generated formulations for the one that ranks anchor responses highest. In experiments on a hydrological calibration benchmark and five antenna design tasks, SHA-PF reaches requirement-satisfying designs in markedly fewer evaluations, succeeding on all five antenna tasks where most expert-based and direct-LLM baselines run out of budget.

Core claim

The paper claims that, among valid formulations of the same design requirement, the ones that give higher priority to rarely observed satisfaction states with dominance support induce search trajectories that reach the requirement-satisfying region in fewer expensive simulator calls. To make this actionable, SHA-PF converts each candidate formulation into an executable objective-plus-constraints program, scores it by the fraction of anchor-vs-other pairs in which the anchor response receives higher priority, and evolves a population of formulations through LLM-based generation, repair, and four variation operators. The authors report that their discovered formulations require substantially f

What carries the argument

The anchor-state formulation objective (Eq. 7) is the load-bearing mechanism: select a non-dominated satisfaction state z_anc with minimal h(z) = occurrence_count(z) / (number_of_initial_points_dominated_by_z + epsilon), then score each candidate formulation by the proportion of pairs (anchor response, other response) in which the anchor ranks higher. This objective turns the abstract idea of 'search hardness' into a concrete fitness for formulation search, guiding the LLM-based evolutionary loop toward formulations that encourage the optimizer to follow hard but promising trajectories.

Load-bearing premise

The load-bearing premise is that 100 space-filling evaluations expose the search-hardness structure of the whole design space; the paper itself concedes that when these samples are sparse or miss hard yet promising regions, anchor selection—and the formulation objective built on it—is misdirected.

What would settle it

Run SHA-PF on a benchmark where the initial 100 evaluations are deliberately constrained to a region far from the known satisfying designs, so no rare promising state is observed. If SHA-PF then fails to reach the requirement within budget while a formulation that merely restates the thresholds succeeds, the claim that initial data reveal progress potential is refuted.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Formulation choice becomes a first-class lever for simulation budget: fixing the optimizer, different valid formulations of the same requirement produce materially different search costs.
  • The initial space-filling evaluations can be reused for formulation search before spending the main optimization budget, so the cost of finding a good formulation is mostly the initial samples themselves.
  • Formulations discovered by SHA-PF transfer across optimizers, suggesting the hardness signal is tied to the landscape rather than to one search algorithm.
  • Search hardness and design-intent alignment become separate, jointly necessary evaluation axes for automatic problem formulation in expensive design.
  • In the antenna benchmarks, the method succeeds where threshold-restating expert and direct-LLM formulations fail, indicating that graded, guidance-shaped formulations matter more than merely encoding the final acceptance criteria.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the finding generalizes, the same anchor-state scoring could be applied to any expensive black-box design setting with natural-language requirements, such as materials discovery or drug design, where the first batch of evaluations is already available.
  • The binary satisfaction state discards margin-to-threshold information; a graded version of the state could make anchor selection smoother, though it might also change which regions count as rare.
  • The paper's proposed iterative anchor discovery—using a few extra evaluations to update satisfaction states during formulation search—is a natural next test and could reduce the sensitivity to the initial sample placement.
  • The rarity-plus-dominance heuristic is one concrete proxy for progress potential; comparing it against surrogate-uncertainty-based or gradient-based anchors on the same benchmarks would clarify what exactly makes a state 'promising.'

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes SHA-PF, a framework that uses an LLM to search over alternative optimization formulations for expensive simulation-driven design. Given an initial set of 100 evaluations D_init, SHA-PF (i) computes a search-hardness score h(z) for each observed binary satisfaction state (Eq. 4), (ii) selects an anchor state z_anc by minimizing h(z) over non-dominated states (Eq. 5), and (iii) scores candidate formulations by the fraction of anchor-vs-other response pairs in which the anchor response ranks higher under the formulation-induced ordering (Eq. 7). The formulations are generated, repaired, and evolved using LLM prompts. The paper reports experiments on the HBV hydrological benchmark and five antenna design tasks, claiming that SHA-PF finds requirement-satisfying designs with significantly fewer evaluations than expert-written or directly LLM-generated formulations.

Significance. If the central claim were established, SHA-PF would be a useful contribution: it addresses a real gap in automated problem formulation by considering search efficiency rather than only requirement alignment. The method is clearly described, the generation/repair/refinement pipeline is sensible, and the inclusion of cross-optimizer and LLM-backbone sensitivity analyses is a strength. However, the empirical evidence as presented does not support the advertised 'significantly fewer evaluations' claim. The results are single runs, the formulation objective is optimized on the same data used to define the benchmark's success, and the reported time budgets are internally inconsistent. The central finding — that prioritizing rare, promising anchor states in the initial data yields formulations that generalize to more efficient search — is asserted rather than demonstrated on held-out data. With additional experiments involving resampling and out-of-sample evaluation, the claim could become credible, but the current version is not.

major comments (4)
  1. [Main Results, Tables 1–2] Every cell in Tables 1 and 2 is a single run, with no variance, no repetitions, and no statistical test. The abstract and text repeatedly say 'significantly fewer evaluations' (e.g., abstract; §Main Results, 'SHA-PF succeeds on all five antenna tasks'), but Table 1 provides only point estimates. Since SHA-PF involves stochastic LLM generation/repair and random initial samples, and the baselines are also stochastic, single runs cannot support a significance claim. Please report multiple independent repetitions with different D_init and LLM sampling, and provide medians, IQRs, and a paired test or effect-size measure.
  2. [Formulation Search Objective, Eq. (7)] The formulation search objective Eq. (7) maximizes anchor-vs-other priority computed on D_init, and the anchor z_anc is itself selected from the same D_init via Eqs. (4)–(5). Moreover, Table 1's caption confirms that the final expensive search for all methods also starts from the same D_init ('excluding the initial evaluations'). Thus a formulation that overfits D_init can score high on Eq. (7) and simultaneously bias the starting population of the final search, without providing any general search-efficiency benefit. The paper's own Conclusion admits that anchor selection 'may be less reliable when these samples are sparse or miss hard yet promising regions.' This is exactly the regime of 100 samples in 13–23 dimensional antenna problems. Please add a held-out or cross-validated protocol: select the formulation on D_init_train and evaluate it on a freshly sampled D_init_test, or repeat
  3. [Experimental Setup and Table 3] There is an internal inconsistency in the evaluation budget. The main text states 'For each antenna design task, the budget is 300 HFSS simulations,' but Table 3's note says the sensitivity experiments use 'a budget of 200 evaluations.' Additionally, the reported time costs in Table 1 cannot be reconciled with the average per-simulation times in Table 8. For example, HAE has an average evaluation time of 532 s, so 66 evaluations would require about 9.7 h, yet Table 1 reports 2.7 h for SHA-PF (SCBO); for CSL, 10 evaluations at 963 s would be about 2.7 h, but Table 1 reports 2.1 h. The discrepancy factor varies across tasks, suggesting unreported parallelism or a different notion of time. Please clarify the evaluation budget and define exactly how 'hours' is measured; otherwise the time savings claims are not reproducible.
  4. [Finding: Initial Data Reveal Search Hardness; Ablation Study] The paper's central finding — 'a formulation is more likely to guide efficient search when it prioritizes rare samples with greater progress potential' — is embedded in the method rather than tested. The ablation replacing anchor selection with random anchor selection (Table 2) compares against the proposed selection on the same D_init, but it does not test alternative hardness definitions or demonstrate that the specific h(z) criterion is the reason for success. To make the claim load-bearing, please add a controlled comparison on held-out D_init: e.g., select formulations by maximizing Eq. (7) with the proposed anchor versus with a random anchor versus with rarity-only or progress-only criteria, and show that the proposed anchor yields better search efficiency on fresh data, not just on the D_init used for selection.
minor comments (5)
  1. [Cross-Optimizer Evaluation] The claim that SHA-PF 'is not tied to a specific optimizer' would be stronger if the baseline formulations were also evaluated with DSI and CEBO, since the current comparison only shows that one SHA-PF formulation works with SCBO, DSI, and CEBO to differing degrees. Without baseline results for the same optimizers, the conclusion that 'optimizer choice is important but not the main reason for the performance gain' is not directly supported.
  2. [Equation (4) notation] In Eq. (4), the denominator notation |{i|z≻z_i}| is slightly ambiguous because z_i ranges over observed states; writing |{i ∈ {1,...,N} : z ≻ z_i}| would be clearer.
  3. [Section 3.3, Φ_R vs. Φ̂_R] The text defines Φ_R as the set of requirement-aligned formulations, but the repair step only 'aims to move' the candidate toward Φ_R, so the actual candidate set Φ̂_R used in Eq. (7) may contain formulations that are not truly requirement-aligned. Please clarify this mismatch and state whether any post-hoc verification is performed before the final optimization.
  4. [Figure 5] In Figure 5(a), the x-axis appears to end at generation 12, while the text and Appendix A use G_max = 25. If the plot is truncated, this should be stated; otherwise the curves should extend to G_max.
  5. [Language] There are several typos and awkward phrasings, e.g., 'The formulation is selected using same initial data that seed the reported' (if present in the final version), and 'allowing more generations provides additional opportunities to improve such populations and reach the maximum objective value.' A light editorial pass would improve readability.

Circularity Check

0 steps flagged

No significant circularity; the formulation-search objective is fitted on D_init, but the reported efficiency metric is measured on new expensive simulations external to D_init.

full rationale

The paper's chain is: initial samples D_init are used to derive satisfaction states, a hardness score h(z) (Eq. 4), an anchor state (Eq. 5), and a formulation-search objective (Eq. 7) that maximizes the fraction of anchor-vs-other pairs ranked in the anchor's favor. This is a deliberate design objective, not a veiled restatement of the reported outcome. The reported claim is that the resulting formulation reduces the number of additional expensive evaluations required to reach a requirement-satisfying design (Table 1). That metric is external to D_init: it is measured by running SCBO (or DSI/CEBO) on the simulator after the initial 100 samples, and the costs exclude those 100 evaluations. Therefore Eq. 7 does not equal the evaluation metric by construction. The 'finding' that rare, dominance-supporting states are valuable is an asserted heuristic with a stated limitation in the Conclusion ('anchor selection may be less reliable when these samples are sparse or miss hard yet promising regions'); that is a validity and generalization concern, not circularity. The only self-citation (Li et al. 2026 for APF) appears in Related Work as background and does not supply the load-bearing argument. Accordingly, no circular step is established, and the paper is best scored at the minor self-citation level rather than at a level implying that the central result reduces to its inputs.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

The method introduces no new physical entities. Its scientific load is carried by the anchor-state heuristic, the binary satisfaction-state representation, and the LLM judge; all are unproven modeling choices that depend on the initial sample set. The free parameters include the hand-set evolutionary settings that Appendix A shows are consequential for success.

free parameters (3)
  • Initial evaluation count N = 100
    Chosen for all benchmarks; no sensitivity analysis is given, and the paper's own conclusion states anchor selection may be unreliable when samples are sparse.
  • Evolutionary parameters P and G_max = 5 and 25
    Appendix A shows P=3 and P=10 fail on HSE and G_max=3 fails, so the main results depend on this hand-chosen default; P=7 and G_max=10 also succeed with different costs.
  • Weights inside discovered formulations = e.g., HSE obj coefficients 0.7, 0.3, 0.6, 0.4, 0.5, 0.8, 0.5; HBV obj weights 20, 10, 2, 1
    These coefficients are generated and evolved by the LLM on D_init and are not derived from first principles. They shape the objective/constraint landscape and therefore the reported search costs.
axioms (4)
  • domain assumption 100 initial space-filling evaluations reveal the search-hardness structure of the whole design space
    Used throughout the Method section; the authors explicitly concede in the Conclusion that anchor selection can be unreliable when samples are sparse or miss hard regions.
  • ad hoc to paper Pareto dominance over binary satisfaction states measures progress potential
    Eq. (4) defines h(z) using the count of dominated states as progress potential; no theoretical or independent empirical justification is given for this specific form.
  • domain assumption The LLM judge correctly maps response curves to binary satisfaction states
    Appendix B.1 relies on an LLM judge to produce compliance vectors; no accuracy validation is reported, yet these vectors determine the anchor and all formulation fitness scores.
  • domain assumption Prioritizing the anchor state under the formulation's ranking transfers to lower search cost with SCBO
    The cross-optimizer and ablation experiments support this empirically, but it is not derived; the formulation search objective (Eq. 7) is a proxy, not a proof.

pith-pipeline@v1.3.0-alltime-deepseek · 26287 in / 10175 out tokens · 122989 ms · 2026-08-01T08:05:54.623195+00:00 · methodology

0 comments
read the original abstract

Expensive simulation-driven design is widely used in engineering to identify requirement-satisfying designs with as few high-fidelity simulations as possible. Most existing efforts address this challenge by improving optimization algorithms under fixed formulations, yet the formulation itself shapes the search landscape by defining the objectives and constraints optimized by the solver. Recent LLM-based automatic problem formulation methods generate formulations from natural-language requirements, but they mainly focus on design-intent alignment and overlook whether the formulation induces an efficient search process. To address this limitation, we propose SHA-PF, a search hardness-aware LLM-based problem formulation framework. We find that a formulation is more likely to guide efficient search when it prioritizes rare samples with greater progress potential. Based on this finding, SHA-PF defines a formulation search objective guided by search hardness, scoring each candidate formulation according to the priority. SHA-PF then searches the formulation space under this objective through LLM-based generation, repair, and evolutionary refinement. Experiments on the real-world multi-objective benchmark and five expensive antenna design benchmarks show that the formulations discovered by SHA-PF require significantly fewer evaluations to reach the design requirements than other baselines.

Figures

Figures reproduced from arXiv: 2607.21220 by Bing Xue, Handing Wang, Mengjie Zhang, Yuchen Li.

Figure 1
Figure 1. Figure 1: Comparison of search paths under different prob [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of search hardness revealed by the initial [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of feasibility convergence and best [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of formulation methods across five antenna design tasks. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Impact of LLM backbone on formulation search [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Convergence of the best-so-far formulation search [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

66 extracted references · 10 canonical work pages · 1 internal anchor

  1. [1]

    doi:10.48550/arXiv.2412.08905 , urldate =

    Phi-4. doi:10.48550/arXiv.2412.08905 , urldate =. arXiv , file =:2412.08905 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2412.08905

  2. [2]

    Proceedings of the 41st

    Ahmaditeshnizi, Ali and Gao, Wenzhi and Udell, Madeleine , year = 2024, month = jul, pages =. Proceedings of the 41st

    2024

  3. [3]

    Integration , volume =

    Analog Circuits Optimization Based on Evolutionary Computation Techniques , author =. Integration , volume =. doi:10.1016/j.vlsi.2009.09.001 , urldate =

    work page doi:10.1016/j.vlsi.2009.09.001 2009

  4. [4]

    Chen, Yitian and Xia, Jingfan and Shao, Siyu and Ge, Dongdong and Ye, Yinyu , year = 2025, month = may, number =. Solver-. doi:10.48550/arXiv.2505.11792 , urldate =. arXiv , keywords =:2505.11792 , primaryclass =

    doi:10.48550/arxiv.2505.11792 2025

  5. [5]

    arXiv , file =:2412.19437 , primaryclass =

    doi:10.48550/arXiv.2412.19437 , urldate =. arXiv , file =:2412.19437 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2412.19437

  6. [6]

    Scalable

    Eriksson, David and Poloczek, Matthias , year = 2021, month = mar, pages =. Scalable. Proceedings of

    2021

  7. [7]

    doi:10.48550/arXiv.2407.21783 , urldate =

    The. doi:10.48550/arXiv.2407.21783 , urldate =. arXiv , file =:2407.21783 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2407.21783

  8. [8]

    doi:10.1287/opre.2024.1233 , urldate =

    Huang, Chenyu and Tang, Zhengyang and Hu, Shixi and Jiang, Ruoqing and Zheng, Xin and Ge, Dongdong and Wang, Benyou and Wang, Zizhuo , year = 2025, month = may, journal =. doi:10.1287/opre.2024.1233 , urldate =

    arXiv 2025

  9. [9]

    Jiang, Caigao and Shu, Xiang and Qian, Hong and Lu, Xingyu and Zhou, Jun and Zhou, Aimin and Yu, Yang , year = 2024, month = oct, urldate =. The

    2024

  10. [10]

    doi:10.48550/arXiv.2310.06825 , urldate =

    Mistral. doi:10.48550/arXiv.2310.06825 , urldate =. arXiv , file =:2310.06825 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2310.06825

  11. [11]

    LLM-based Automated Theorem Proving Hinges on Scalable Synthetic Data Generation

    Lai, Junyu and Zhang, Jiakun and Xu, Shuo and Chen, Taolue and Wang, Zihang and Yang, Yao and Zhang, Jiarui and Cao, Chun and Xu, Jingwei , year = 2025, month = may, number =. doi:10.48550/arXiv.2505.12031 , urldate =. arXiv , keywords =:2505.12031 , primaryclass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2505.12031 2025

  12. [12]

    Lei, Shiwen and Yang, Yaohui and Hu, Haoquan and Zhao, Zhiqin and Chen, Bo and Qiu, Xiangdong , year = 2019, month = mar, journal =. Power. doi:10.1109/TAP.2018.2884814 , urldate =

    arXiv 2019

  13. [14]

    Progress in Aerospace Sciences , volume =

    Machine Learning in Aerodynamic Shape Optimization , author =. Progress in Aerospace Sciences , volume =. doi:10.1016/j.paerosci.2022.100849 , urldate =

    arXiv 2022

  14. [15]

    Mechanical Systems and Signal Processing , volume =

    Study of Structural Optimization Design on a Certain Vehicle Body-in-White Based on Static Performance and Modal Analysis , author =. Mechanical Systems and Signal Processing , volume =. doi:10.1016/j.ymssp.2019.106405 , urldate =

    arXiv 2019

  15. [16]

    doi:10.48550/arXiv.2402.13064 , urldate =

    Synthetic. doi:10.48550/arXiv.2402.13064 , urldate =. arXiv , keywords =:2402.13064 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2402.13064

  16. [17]

    doi:10.48550/arXiv.2404.07503 , urldate =

    Best. doi:10.48550/arXiv.2404.07503 , urldate =. arXiv , file =:2404.07503 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2404.07503

  17. [18]

    Long, Lin and Wang, Rui and Xiao, Ruixuan and Zhao, Junbo and Ding, Xiao and Chen, Gang and Wang, Haobo , editor =. On. Findings of the. doi:10.18653/v1/2024.findings-acl.658 , urldate =

    doi:10.18653/v1/2024.findings-acl.658 2024

  18. [19]

    doi:10.48550/arXiv.2502.11102 , urldate =

    Lu, Hongliang and Xie, Zhonglin and Wu, Yaoyu and Ren, Can and Chen, Yuxuan and Wen, Zaiwen , year = 2025, month = feb, number =. doi:10.48550/arXiv.2502.11102 , urldate =. arXiv , keywords =:2502.11102 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2502.11102 2025

  19. [20]

    Luo, Ziyang and Xu, Can and Zhao, Pu and Sun, Qingfeng and Geng, Xiubo and Hu, Wenxiang and Tao, Chongyang and Ma, Jing and Lin, Qingwei and Jiang, Daxin , year = 2023, month = oct, urldate =. The

    2023

  20. [21]

    Lyu, Zhoujie and Kenway, Gaetan K. W. and Martins, Joaquim R. R. A. , year = 2015, month = apr, journal =. Aerodynamic. doi:10.2514/1.J053318 , urldate =

    work page doi:10.2514/1.j053318 2015

  21. [22]

    doi:10.48550/arXiv.2403.01131 , urldate =

    Ma, Zeyuan and Guo, Hongshu and Chen, Jiacheng and Peng, Guojun and Cao, Zhiguang and Ma, Yining and Gong, Yue-Jiao , year = 2024, month = mar, number =. doi:10.48550/arXiv.2403.01131 , urldate =. arXiv , keywords =:2403.01131 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2403.01131 2024

  22. [23]

    arXiv , file =:2407.03502 , primaryclass =

    doi:10.48550/arXiv.2407.03502 , urldate =. arXiv , file =:2407.03502 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2407.03502

  23. [24]

    arXiv , file =:2303.08774 , primaryclass =

    doi:10.48550/arXiv.2303.08774 , urldate =. arXiv , file =:2303.08774 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2303.08774

  24. [25]

    doi:10.48550/arXiv.2412.15115 , urldate =

    Qwen2.5. doi:10.48550/arXiv.2412.15115 , urldate =. arXiv , keywords =:2412.15115 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2412.15115

  25. [26]

    Microwave Journal , volume =

    Electromagnetic. Microwave Journal , volume =

  26. [27]

    and Sivaramakrishnan, R

    Shanmugasundar, G. and Sivaramakrishnan, R. and Meganathan, S. and Balasubramani, S. , year = 2019, journal =. Structural. doi:10.1016/j.matpr.2019.05.231 , urldate =

    work page doi:10.1016/j.matpr.2019.05.231 2019

  27. [28]

    doi:10.48550/arXiv.2403.05530 , urldate =

    Gemini 1.5:. doi:10.48550/arXiv.2403.05530 , urldate =. arXiv , keywords =:2403.05530 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2403.05530

  28. [29]

    doi:10.48550/arXiv.2312.11805 , urldate =

    Gemini:. doi:10.48550/arXiv.2312.11805 , urldate =. arXiv , file =:2312.11805 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2312.11805

  29. [30]

    and Khashabi, Daniel and Hajishirzi, Hannaneh , year = 2023, month = may, number =

    Wang, Yizhong and Kordi, Yeganeh and Mishra, Swaroop and Liu, Alisa and Smith, Noah A. and Khashabi, Daniel and Hajishirzi, Hannaneh , year = 2023, month = may, number =. Self-. doi:10.48550/arXiv.2212.10560 , urldate =. 2212.10560 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2212.10560 2023

  30. [31]

    doi:10.48550/arXiv.2405.14333 , urldate =

    Xin, Huajian and Guo, Daya and Shao, Zhihong and Ren, Zhizhou and Zhu, Qihao and Liu, Bo and Ruan, Chong and Li, Wenda and Liang, Xiaodan , year = 2024, month = may, number =. doi:10.48550/arXiv.2405.14333 , urldate =. arXiv , keywords =:2405.14333 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2405.14333 2024

  31. [32]

    Xu, Zhangchen and Jiang, Fengqing and Niu, Luyao and Deng, Yuntian and Poovendran, Radha and Choi, Yejin and Lin, Bill Yuchen , year = 2024, month = oct, urldate =. Magpie:. The

    2024

  32. [33]

    , year = 2022, month = may, number =

    Zelikman, Eric and Wu, Yuhuai and Mu, Jesse and Goodman, Noah D. , year = 2022, month = may, number =. doi:10.48550/arXiv.2203.14465 , urldate =. 2203.14465 , primaryclass =

    Pith/arXiv arXiv doi:10.48550/arxiv.2203.14465 2022

  33. [34]

    Journal of Optimization Theory and Applications , year = 1974, volume =

    Cone Convexity, Cone Extreme Points, and Nondominated Solutions in Decision Problems with Multiobjectives , author =. Journal of Optimization Theory and Applications , year = 1974, volume =. doi:10.1007/BF00932614 , issn =

    work page doi:10.1007/bf00932614 1974

  34. [35]

    International Journal of RF and Microwave Computer-Aided Engineering , year = 2020, volume =

    A Review on the Design and Optimization of Antennas Using Machine Learning Algorithms and Techniques , author =. International Journal of RF and Microwave Computer-Aided Engineering , year = 2020, volume =. doi:10.1002/mmce.22356 , issn =

    work page doi:10.1002/mmce.22356 2020

  35. [36]

    2025 , url =

    Guy Tevet and Sigal Raab and Setareh Cohan and Daniele Reda and Zhengyi Luo and Xue Bin Peng and Amit Haim Bermano and Michiel van de Panne , booktitle =. 2025 , url =

    2025

  36. [37]

    Advances in Neural Information Processing Systems , year = 2024, volume =

    Scalable and Effective Arithmetic Tree Generation for Adder and Multiplier Designs , author =. Advances in Neural Information Processing Systems , year = 2024, volume =. doi:10.52202/079017-4416 , editor =

    work page doi:10.52202/079017-4416 2024

  37. [38]

    2025 IEEE-APS Topical Conference on Antennas and Propagation in Wireless Communications (APWC) , year =

    Surrogate Modeling with Hybrid CNN-RNN and GAN Architectures for Antenna Designs , author =. 2025 IEEE-APS Topical Conference on Antennas and Propagation in Wireless Communications (APWC) , year =. doi:10.1109/APWC65665.2025.11190408 , keywords =

    arXiv 2025

  38. [39]

    AIAA SCITECH 2025 Forum , year =

    Multi-Fidelity Surrogate Modeling for Aerodynamics Using Sage , author =. AIAA SCITECH 2025 Forum , year =

    2025

  39. [40]

    Proceedings of the AAAI Conference on Artificial Intelligence , year =

    Paretohqd: Fast offline multiobjective alignment of large language models using pareto high-quality data , author =. Proceedings of the AAAI Conference on Artificial Intelligence , year =

  40. [41]

    Gu, Haoran and Wang, Handing and Mei, Yi and Zhang, Mengjie and Jin, Yaochu , year = 2026, archiveprefix =. One. 2505.07167 , month = jan, number =

    arXiv 2026

  41. [42]

    2026 , journal =

    Solver-Independent Automated Problem Formulation via LLMs for High-Cost Simulation-Driven Design , author =. 2026 , journal =

    2026

  42. [43]

    Zhang, Yisong and Cheng, Ran and Yi, Guoxing and Tan, Kay Chen , year = 2025, abstract =. A. arXiv , doi =:2509.08269 , keywords =

    Pith/arXiv arXiv 2025

  43. [44]

    Jin, Yaochu and Wang, Handing and Chugh, Tinkle and Guo, Dan and Miettinen, Kaisa , journal =. Data-. doi:10.1109/TEVC.2018.2869001 , issn =

    arXiv 2018

  44. [45]

    Xu, Yiteng and Guo, Qingbin and Gu, Haoran and Wang, Handing and Yan, Junchi , journal =. Self-. doi:10.1109/TEVC.2026.3662703 , issn =

    arXiv 2026

  45. [46]

    Wang, Xilu and Jin, Yaochu and Schmitt, Sebastian and Olhofer, Markus , journal =. Recent. doi:10.1145/3582078 , issn =

    doi:10.1145/3582078

  46. [47]

    Liu, Songbai and Lin, Qiuzhen and Li, Jianqiang and Tan, Kay Chen , journal =. A. doi:10.1109/TEVC.2023.3250350 , issn =

    arXiv 2023

  47. [48]

    Liu, Bo and Zhang, Qingfu and Gielen, Georges G. E. , journal =. A. doi:10.1109/TEVC.2013.2248012 , issn =

    arXiv 2013

  48. [49]

    Chugh, Tinkle and Jin, Yaochu and Miettinen, Kaisa and Hakanen, Jussi and Sindhya, Karthik , journal =. A. doi:10.1109/TEVC.2016.2622301 , issn =

    arXiv 2016

  49. [50]

    Scalable Global Optimization via Local

    Eriksson, David and Pearce, Michael and Gardner, Jacob R and Turner, Ryan and Poloczek, Matthias , booktitle =. Scalable Global Optimization via Local

  50. [51]

    Scalable

    Eriksson, David and Poloczek, Matthias , booktitle =. Scalable

  51. [52]

    Differentiable Expected Hypervolume Improvement for Parallel Multi-Objective

    Daulton, Samuel and Balandat, Maximilian and Bakshy, Eytan , booktitle =. Differentiable Expected Hypervolume Improvement for Parallel Multi-Objective

  52. [53]

    Parallel

    Daulton, Samuel and Balandat, Maximilian and Bakshy, Eytan , booktitle =. Parallel

  53. [54]

    , journal =

    Alam, Khairul and Ray, Tapabrata and Anavatti, Sreenatha G. , journal =. Design. doi:10.1109/TSMC.2015.2390592 , issn =

    arXiv 2015

  54. [55]

    Augmenting

    Ramamonjison, Rindra and Li, Haley and Yu, Timothy and He, Shiqi and Rengan, Vishnu and. Augmenting. Proceedings of the 2022. doi:10.18653/v1/2022.emnlp-industry.4 , editor =

    doi:10.18653/v1/2022.emnlp-industry.4 2022

  55. [56]

    Zhang, Jihai and Wang, Wei and Guo, Siyan and Wang, Li and Lin, Fangquan and Yang, Cheng and Yin, Wotao , booktitle =. Solving. doi:10.18653/v1/2024.naacl-industry.42 , editor =

    work page doi:10.18653/v1/2024.naacl-industry.42 2024

  56. [57]

    Hill, Cristina and Suh, Jina and Sarrafzadeh, Bahareh , journal =

    Lawless, Connor and Schoeffer, Jakob and Le, Lindy and Rowan, Kael and Sen, Shilad and St. Hill, Cristina and Suh, Jina and Sarrafzadeh, Bahareh , journal =. ``. doi:10.1145/3685053 , issn =

    work page doi:10.1145/3685053

  57. [58]

    Lu, Hongliang and Xie, Zhonglin and Wu, Yaoyu and Ren, Can and Chen, Yuxuan and Wen, Zaiwen , booktitle =

  58. [59]

    doi:10.48550/arXiv.2407.09887 , eprint =

    Yang, Zhicheng and Wang, Yiwei and Huang, Yinya and Guo, Zhijiang and Shi, Wei and Han, Xiongwei and Feng, Liang and Song, Linqi and Liang, Xiaodan and Tang, Jing , year = 2025, archiveprefix =. doi:10.48550/arXiv.2407.09887 , eprint =

    Pith/arXiv arXiv doi:10.48550/arxiv.2407.09887 2025

  59. [60]

    Reed, P. M. and Hadka, D. and Herman, J. D. and Kasprzyk, J. R. and Kollat, J. B. , journal =. Evolutionary Multiobjective Optimization in Water Resources:. doi:10.1016/j.advwatres.2012.01.005 , issn =

    work page doi:10.1016/j.advwatres.2012.01.005 2012

  60. [61]

    Design of

    Guo, Junlei and Chen, Yongpin and Yang, Deqiang and Sun, Kai and Pan, Jin and Liu, Sihao , journal =. Design of. doi:10.1109/TAP.2023.3331253 , issn =

    arXiv 2023

  61. [62]

    Design of

    Hu, Kun-Zhi and Tang, Ming-Chun and Li, Dajiang and Wang, Yang and Li, Mei , journal =. Design of. doi:10.1109/TAP.2019.2938574 , issn =

    arXiv 2019

  62. [63]

    Liu, Shuxuan and Wang, Zhan and Dong, Yuandan , journal =. A. doi:10.1109/LAWP.2022.3224845 , issn =

    arXiv 2022

  63. [64]

    Yuan, Hang and Chen, Fu-Chang and Chu, Qing-Xin , journal =. A. doi:10.1109/TAP.2022.3177494 , issn =

    arXiv 2022

  64. [65]

    Liang, Gen-Zhu and Chen, Fu-Chang and Yuan, Hang and Xiang, Kai-Ran and Chu, Qing-Xin , journal =. A. doi:10.1109/TAP.2021.3098563 , issn =

    arXiv 2021

  65. [66]

    Song, Zhenshou and Wang, Handing and Jin, Yaochu , journal =. A. doi:10.1109/TSMC.2023.3281822 , issn =

    arXiv 2023

  66. [67]

    Constrained

    Liu, Jiao and Wang, Yong and Sun, Guangyong and Pang, Tong , journal =. Constrained. doi:10.1109/TSMC.2024.3504728 , issn =

    arXiv 2024