Pith. sign in

REVIEW 2 major objections 61 references

Representing acoustic metamaterials as sequences lets a generative model solve inverse design for broadband responses more accurately than image or template approaches.

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 · grok-4.3

2026-06-27 15:13 UTC pith:OOC6TPST

load-bearing objection Sequence representation plus external-solver RL is a sensible step for AMM inverse design, but the 45% error reduction sits on thin reported evidence. the 2 major comments →

arxiv 2606.09266 v1 pith:OOC6TPST submitted 2026-06-08 cs.SD cs.AI

Physics-Guided Sequence-Based Generative Framework for Acoustic Metamaterial Inverse Design

Yijie Li , Jiahao Xu , Ching-Chih Tsao , Lili Qiu , Jingxian Wang This is my paper
classification cs.SD cs.AI
keywords acoustic metamaterialsinverse designsequence modelinggenerative designphysics-guided learningbroadband responsereinforcement learning
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 establishes that acoustic metamaterial inverse design for broadband targets can be reframed as a sequence-to-sequence task when structures are encoded as ordered sequences rather than pixel grids or fixed templates. This encoding preserves exact geometry and connectivity while allowing supervised pretraining followed by physics-solver-guided reinforcement learning to resolve the one-to-many mapping from response to structure. The resulting MetaSeq framework is shown to cut response error by 45 percent relative to the strongest of five baselines when evaluated against full-wave COMSOL simulations. A sympathetic reader would care because dispersion makes independent control of adjacent frequency bands difficult; a method that respects connectivity and geometry directly could therefore produce usable broadband absorbers or filters without extensive manual tuning.

Core claim

MetaSeq introduces a language that represents each acoustic metamaterial as a structured sequence, rather than as a pixel grid or fixed template. This representation preserves precise geometry, explicitly encodes connectivity, and casts inverse design as a sequence-to-sequence task from target response to structure sequence. MetaSeq further constructs a balanced, high-fidelity dataset with efficient calibration and complexity-based sampling, then combines supervised pretraining with reinforcement learning fine-tuning guided by a physics-based solver and validity checker.

What carries the argument

The sequence language for AMMs that preserves precise geometry and explicitly encodes connectivity, turning inverse design into a sequence-to-sequence generation task.

Load-bearing premise

The sequence language fully preserves the geometric precision and structural connectivity required for accurate acoustic performance.

What would settle it

Fabricate the top MetaSeq-generated structure and measure its actual transmission or reflection spectrum across the target band; if measured error exceeds the reported 45 percent reduction relative to the best baseline, the simulation-to-reality transfer claim fails.

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

If this is right

  • Inverse design becomes possible directly from broadband target responses without being limited to predefined templates.
  • Structural connectivity is maintained explicitly, avoiding the loss of acoustic paths that occurs in image-based encodings.
  • Physics-solver feedback during reinforcement learning improves consistency across multiple frequencies simultaneously.
  • Complexity-based sampling produces training data that better covers the space of manufacturable geometries.

Where Pith is reading between the lines

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

  • The same sequence representation could be tested on other wave phenomena such as elastic or electromagnetic metamaterials where connectivity likewise governs performance.
  • Extending the sequence vocabulary to three-dimensional or multi-material stacks would test whether the approach scales beyond planar designs.
  • Because the method separates representation from the solver, it could be paired with faster surrogate models to reduce the cost of the RL fine-tuning stage.

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

2 major / 0 minor

Summary. The paper introduces MetaSeq, a physics-guided sequence-based generative framework for inverse design of acoustic metamaterials targeting broadband responses. It proposes a structured sequence language to represent AMM geometries (claimed to preserve precise geometry and explicit connectivity better than image or template methods), constructs a balanced high-fidelity dataset via calibration and complexity-based sampling, and combines supervised pretraining with RL fine-tuning that uses an external physics solver and validity checker for rewards. The central empirical claim is that MetaSeq achieves a 45% reduction in response error relative to the best of five baselines, with validation against COMSOL simulations.

Significance. If the performance result and representation fidelity hold under detailed scrutiny, the work would offer a meaningful advance in handling acoustic dispersion for broadband AMM design by moving beyond constrained templates or lossy image encodings. The integration of physics-based guidance in the RL stage is a positive methodological feature, though the absence of supporting evaluation details currently limits assessment of its broader impact on the field.

major comments (2)
  1. [Abstract] Abstract: the headline claim that MetaSeq reduces response error by 45% over the best baseline is presented without error bars, dataset size, baseline implementation details, or description of how the 45% figure was computed. This information is load-bearing for the central performance claim and prevents checking for post-hoc choices or evaluation leakage.
  2. [Abstract] Abstract (representation introduction): the assertion that the sequence language 'preserves precise geometry, explicitly encodes connectivity, and overcomes the limitations of image-based or template-based methods' lacks any quantitative bound on representation error, enumeration of unrepresentable structures, or ablation removing the connectivity encoding. This is load-bearing for attributing the reported error reduction to the proposed method rather than to an incomplete search space.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment point by point below and agree that the abstract requires strengthening to better support the central claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the headline claim that MetaSeq reduces response error by 45% over the best baseline is presented without error bars, dataset size, baseline implementation details, or description of how the 45% figure was computed. This information is load-bearing for the central performance claim and prevents checking for post-hoc choices or evaluation leakage.

    Authors: We agree that the abstract as currently written does not contain these supporting details, which limits immediate verifiability of the headline result. The body of the manuscript reports the evaluation protocol, but to make the abstract self-contained we will revise it to state the dataset size, note that error bars are computed over multiple independent runs, briefly indicate how baselines were implemented, and describe the relative-error formula used for the 45% figure. This revision will be incorporated in the next version of the manuscript. revision: yes

  2. Referee: [Abstract] Abstract (representation introduction): the assertion that the sequence language 'preserves precise geometry, explicitly encodes connectivity, and overcomes the limitations of image-based or template-based methods' lacks any quantitative bound on representation error, enumeration of unrepresentable structures, or ablation removing the connectivity encoding. This is load-bearing for attributing the reported error reduction to the proposed method rather than to an incomplete search space.

    Authors: We acknowledge that the abstract presents the representation advantages without accompanying quantitative support or ablation evidence. The sequence encoding is constructed to be exact (zero representation error) for all geometries inside the considered design space, but the manuscript does not currently supply a numerical bound on error outside that space, an explicit enumeration of excluded structures, or a dedicated ablation on connectivity. We will therefore revise the abstract to qualify the claim and will add the requested quantitative discussion, enumeration, and ablation reference to the main text in the revised manuscript. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper introduces a sequence representation for AMMs and trains via supervised pretraining plus RL whose rewards come from an external physics solver and validity checker. The headline 45% error reduction is measured on COMSOL simulations against independent baselines, not re-expressed from quantities fitted inside the training loop or from self-citations. No self-definitional equations, fitted-input predictions, or load-bearing self-citation chains appear in the abstract or described method; the representation is presented as an engineering choice whose fidelity is asserted rather than derived by construction from the performance metric.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review yields no explicit free parameters, axioms, or invented entities beyond the high-level claim that the sequence language and RL loop improve performance; the physics solver is treated as an external oracle whose accuracy is assumed.

pith-pipeline@v0.9.1-grok · 5753 in / 1262 out tokens · 25918 ms · 2026-06-27T15:13:09.108689+00:00 · methodology

0 comments
read the original abstract

Acoustic metamaterial (AMM) inverse design is particularly challenging for broadband target responses due to acoustic dispersion: a structure that matches the desired response at one frequency may deviate at others, and modifying geometry to improve one sub-band often perturbs neighboring sub-bands. Yet existing broadband inverse-design approaches are either constrained by predefined templates, or rely on image representations that fail to preserve the geometric precision and structural connectivity required by acoustic structures. We present MetaSeq, a physics-guided, sequence-based generative framework for acoustic metamaterial inverse design. At its core, MetaSeq introduces a language that represents each AMM as a structured sequence, rather than as a pixel grid or fixed template. This representation preserves precise geometry, explicitly encodes connectivity, and casts inverse design as a sequence-to-sequence task from target response to structure sequence. MetaSeq further constructs a balanced, high-fidelity dataset with efficient calibration and complexity-based sampling. To address the one-to-many nature of inverse design, MetaSeq combines supervised pretraining with reinforcement learning fine-tuning guided by a physics-based solver and validity checker. Extensive evaluations against COMSOL and five baselines show that MetaSeq reduces response error by 45% over the best baseline.

Figures

Figures reproduced from arXiv: 2606.09266 by Ching-Chih Tsao, Jiahao Xu, Jingxian Wang, Lili Qiu, Yijie Li.

Figure 1
Figure 1. Figure 1: MetaSeq casts acoustic metamaterial inverse design as generating a sequence representation of an AMM from a target acoustic response. This work focuses on the inverse design of AMM units, where the goal is to generate a structure1 that realizes a target response, i.e., desired phase and amplitude. This prob￾lem becomes substantially challenging for broadband tar￾gets [13, 20, 33, 42], which require a desir… view at source ↗
Figure 2
Figure 2. Figure 2: Scaling images to reduce training overhead [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Example structural sequence in MetaSeq’s language. 4.2 Grammar The syntax of MetaSeq governs how tokens from the vo￾cabulary are assembled into valid sequences. To define the ordering of the tokens, we adopt a Depth-First Traversal (DFT) strategy [12] to linearize the acoustic tree structure: Every time the traversal encounters a side-branch (e.g., a Helmholtz resonator shunt to a duct), the sequence enter… view at source ↗
Figure 5
Figure 5. Figure 5: Distributions of component counts and se [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: MetaSeq’s Model Architecture dataset ( [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Performance against the training set volume. [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Ablation Performance. presents representative generated responses for diverse inverse￾design objectives, including transmission responses with one, two, or three spectral peaks and wideband coverage, in both the high- and low-frequency bands [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Structural Diversity. nature of inverse design by encouraging multiple distinct structures for the same target response, rather than overfit￾ting to a single reference design, thereby better supporting downstream needs such as fabrication robustness or material usage. As shown in [PITH_FULL_IMAGE:figures/full_fig_p012_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Representative target and generated responses of [PITH_FULL_IMAGE:figures/full_fig_p013_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Illustration of physical solver for acoustics. [PITH_FULL_IMAGE:figures/full_fig_p016_12.png] view at source ↗
Figure 14
Figure 14. Figure 14: Validation loss across checkpoints. E Diffusion Implementation Architecture and Formulation. The generation mod￾ule is designed to produce acoustic metamaterial patterns x ∈ {0, 1} 96×192 that correspond to a specific target acoustic response y. This response is structured as a 100 × 2 matrix covering frequency-dependent transmission, and phase val￾ues, which is subsequently flattened into a 200-dimension… view at source ↗
Figure 15
Figure 15. Figure 15: Hyperparameters selection during training. [PITH_FULL_IMAGE:figures/full_fig_p018_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Performance under balanced dataset and random dataset. performance optimization and constraint satisfaction. We sweep both coefficients (e.g., structural penalty weight and overlap penalty weight) [PITH_FULL_IMAGE:figures/full_fig_p018_16.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

61 extracted references · 3 canonical work pages · 2 internal anchors

  1. [1]

    Inverse design of broadband acoustic metasurfaces for reflective wave- front modulation through the topology optimization method.Applied Acoustics, 204:109247, 2023

    2023

  2. [2]

    Journal of Sound and Vibration, 596:118789, 2025

    Acoustic structure inverse design and optimization using deep learning. Journal of Sound and Vibration, 596:118789, 2025

    2025

  3. [3]

    Performance prediction and inverse design of cylindrical plate-type acoustic metamaterials based on deep learning.Applied Acoustics, 234:110633, 2025

    2025

  4. [4]

    A new inverse design method for sound-absorbing metamaterial based on deep learning.Applied Acoustics, 241:111024, 2026

    2026

  5. [5]

    Amirkulova, Peter Lai, and Diego Soto

    Feruza A. Amirkulova, Peter Lai, and Diego Soto. Design of acoustic metamaterials using conditional wasserstein generative adversarial networks. InProceedings of the International Congress on Acoustics, 2022

    2022

  6. [6]

    Multifunctional metasurface design with a generative adversarial network.Advanced optical materials, 9(5):2001433, 2021

    Sensong An, Bowen Zheng, Hong Tang, Mikhail Y Shalaginov, Li Zhou, Hang Li, Myungkoo Kang, Kathleen A Richardson, Tian Gu, Juejun Hu, et al. Multifunctional metasurface design with a generative adversarial network.Advanced optical materials, 9(5):2001433, 2021

    2021

  7. [7]

    Acous- tic metasurfaces.Nature Reviews Materials, 3:460–472, 2018

    Badreddine Assouar, Bin Liang, Ying Wu, Yong Li, and Yun Jing. Acous- tic metasurfaces.Nature Reviews Materials, 3:460–472, 2018

    2018

  8. [8]

    Bendsøe and Ole Sigmund.Topology Optimization: Theory, Methods, and Applications

    Martin P. Bendsøe and Ole Sigmund.Topology Optimization: Theory, Methods, and Applications. Springer, 2004

    2004

  9. [9]

    Wright, and Sam H

    Eun Bok, Jong Jin Park, Haejin Choi, Chung Kyu Han, Oliver B. Wright, and Sam H. Lee. Metasurface for water-to-air sound transmission. Physical Review Letters, 120(4):044302, 2018

    2018

  10. [10]

    John Wiley & Sons, 2013

    Michel Bruneau.Fundamentals of acoustics. John Wiley & Sons, 2013

    2013

  11. [11]

    Collision detection using axis aligned bounding boxes

    Panpan Cai, Chandrasekaran Indhumathi, Yiyu Cai, Jianmin Zheng, Yi Gong, Teng Sam Lim, and Peng Wong. Collision detection using axis aligned bounding boxes. InSimulations, Serious Games and Their Applications, pages 1–14. Springer, 2013

    2013

  12. [12]

    A unified approach to parallel depth-first traversals of general trees.Information Processing Letters, 38(1):49–55, 1991

    Calvin C-Y Chen, Sajal K Das, and Selim G Akl. A unified approach to parallel depth-first traversals of general trees.Information Processing Letters, 38(1):49–55, 1991

    1991

  13. [13]

    Broadband acoustic focusing by airy-like beams based on acoustic metasurfaces.Journal of Applied Physics, 123(4), 2018

    Di-Chao Chen, Xing-Feng Zhu, Qi Wei, Da-Jian Wu, and Xiao-Jun Liu. Broadband acoustic focusing by airy-like beams based on acoustic metasurfaces.Journal of Applied Physics, 123(4), 2018

    2018

  14. [14]

    Approximation by superpositions of a sigmoidal function.Mathematics of control, signals and systems, 2(4):303–314, 1989

    George Cybenko. Approximation by superpositions of a sigmoidal function.Mathematics of control, signals and systems, 2(4):303–314, 1989

    1989

  15. [15]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. InProceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers), pages 4171– 4186, 2019

    2019

  16. [16]

    Acoustic metamateri- als.MRS bulletin, 33(10):931–934, 2008

    Lee Fok, Muralidhar Ambati, and Xiang Zhang. Acoustic metamateri- als.MRS bulletin, 33(10):931–934, 2008

    2008

  17. [17]

    Adaptive metasurface-based acoustic imaging using joint optimization

    Yongjian Fu, Yongzhao Zhang, Yu Lu, Lili Qiu, Yi-Chao Chen, Yezhou Wang, Mei Wang, Yijie Li, Ju Ren, and Yaoxue Zhang. Adaptive metasurface-based acoustic imaging using joint optimization. InPro- ceedings of the 22nd Annual International Conference on Mobile Systems, Applications and Services, pages 492–504, 2024

    2024

  18. [18]

    Pushing the limits of acoustic spatial perception via incident angle encoding.Pro- ceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 8(2):1–28, 2024

    Yongjian Fu, Yongzhao Zhang, Hao Pan, Yu Lu, Xinyi Li, Lili Chen, Ju Ren, Xiong Li, Xiaosong Zhang, and Yaoxue Zhang. Pushing the limits of acoustic spatial perception via incident angle encoding.Pro- ceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 8(2):1–28, 2024

    2024

  19. [19]

    Linear hinge loss and average margin.Advances in neural information processing systems, 11, 1998

    Claudio Gentile and Manfred KK Warmuth. Linear hinge loss and average margin.Advances in neural information processing systems, 11, 1998

    1998

  20. [20]

    A phase- gradient acoustic metasurface for broadband duct noise attenuation in the presence of flow.International Journal of Mechanical Sciences, 237:107822, 2023

    Jingwen Guo, Renhao Qu, Yi Fang, Wei Yi, and Xin Zhang. A phase- gradient acoustic metasurface for broadband duct noise attenuation in the presence of flow.International Journal of Mechanical Sciences, 237:107822, 2023

    2023

  21. [21]

    Matrix formulation in acoustics: the transfer matrix method

    Iván Herrero Durá, Alejandro Cebrecos Ruiz, Luis Miguel García Raffi, and Vicente Romero-García. Matrix formulation in acoustics: the transfer matrix method. 2019

    2019

  22. [22]

    Adap- tive genetic algorithm for optical metasurfaces design.Scientific reports, 8(1):11040, 2018

    Samad Jafar-Zanjani, Sandeep Inampudi, and Hossein Mosallaei. Adap- tive genetic algorithm for optical metasurfaces design.Scientific reports, 8(1):11040, 2018

    2018

  23. [23]

    Topology optimization for nano-photonics.Laser & Photonics Reviews, 5(2):308–321, 2011

    Jakob Søndergaard Jensen and Ole Sigmund. Topology optimization for nano-photonics.Laser & Photonics Reviews, 5(2):308–321, 2011

    2011

  24. [24]

    Genetic algorithms in engi- neering electromagnetics.IEEE Antennas and propagation Magazine, 39(4):7–21, 1997

    J Michael Johnson and V Rahmat-Samii. Genetic algorithms in engi- neering electromagnetics.IEEE Antennas and propagation Magazine, 39(4):7–21, 1997. Conference’17, July 2017, Washington, DC, USA , Yijie Li, Jiahao Xu, Ching-Chih Tsao, Lili Qiu, and Jingxian Wang

    1997

  25. [25]

    Gelu activation function in deep learning: a com- prehensive mathematical analysis and performance.arXiv preprint arXiv:2305.12073, 2023

    Minhyeok Lee. Gelu activation function in deep learning: a com- prehensive mathematical analysis and performance.arXiv preprint arXiv:2305.12073, 2023

    work page arXiv 2023

  26. [26]

    Binary codes capable of correcting dele- tions, insertions, and reversals

    Vladimir I Levenshtein et al. Binary codes capable of correcting dele- tions, insertions, and reversals. InSoviet physics doklady, volume 10, pages 707–710. Soviet Union, 1966

    1966

  27. [27]

    Cum- mer

    Dennis Li, Lucian Zigoneanu, Bogdan-Ioan Popa, and Steven A. Cum- mer. Design of an acoustic metamaterial lens using genetic algorithms. The Journal of the Acoustical Society of America, 132(4):2823–2833, 2012

    2012

  28. [28]

    On sampling top-k recommendation evaluation

    Dong Li, Ruoming Jin, Jing Gao, and Zhi Liu. On sampling top-k recommendation evaluation. InProceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 2114–2124, 2020

    2020

  29. [29]

    Sound insulation performance of double membrane-type acoustic metamateri- als combined with a helmholtz resonator.Applied Acoustics, 205:109297, 2023

    Hong-Ze Li, Xu-Chang Liu, Qi Liu, Shuang Li, Jin-Shui Yang, Li-Li Tong, Sheng-Bo Shi, Rüdiger Schmidt, and Kai-Uwe Schröder. Sound insulation performance of double membrane-type acoustic metamateri- als combined with a helmholtz resonator.Applied Acoustics, 205:109297, 2023

    2023

  30. [30]

    Mudis: An audio-independent, wide-angle, and leak-free multi-directional speaker

    Yijie Li, Juntao Zhou, Dian Ding, Yi-Chao Chen, Lili Qiu, Jiadi Yu, and Guangtao Xue. Mudis: An audio-independent, wide-angle, and leak-free multi-directional speaker. InProceedings of the 30th Annual International Conference on Mobile Computing and Networking, pages 263–278, 2024

    2024

  31. [31]

    Acoustic metasurface-based perfect absorber with deep subwavelength thickness.Applied Physics Letters, 108(6):063502, 2016

    Yong Li and Badreddine Assouar. Acoustic metasurface-based perfect absorber with deep subwavelength thickness.Applied Physics Letters, 108(6):063502, 2016

    2016

  32. [32]

    Extreme acoustic metamaterial by coiling up space.Physical review letters, 108(11):114301, 2012

    Zixian Liang and Jensen Li. Extreme acoustic metamaterial by coiling up space.Physical review letters, 108(11):114301, 2012

    2012

  33. [33]

    Achromatic transmitted acoustic lens arrayed by cascaded helmholtz resonators.Applied Physics Express, 16(1):014004, 2023

    Cheng Lü, Shuai Tang, Jin-Lei Wu, Yanbo Pei, and Yongyuan Jiang. Achromatic transmitted acoustic lens arrayed by cascaded helmholtz resonators.Applied Physics Express, 16(1):014004, 2023

    2023

  34. [34]

    Deep learning-based design method for acoustic metasurface dual- feature fusion.Materials, 17(9):2166, 2024

    Qiang Lv, Huanlong Zhao, Zhen Huang, Guoqiang Hao, and Wei Chen. Deep learning-based design method for acoustic metasurface dual- feature fusion.Materials, 17(9):2166, 2024

    2024

  35. [35]

    Acoustic metasurface with hybrid resonances.Nature Materials, 13:873–878, 2014

    Guancong Ma, Mengxin Yang, Shu Xiao, Zhaoxiong Yang, and Ping Sheng. Acoustic metasurface with hybrid resonances.Nature Materials, 13:873–878, 2014

    2014

  36. [36]

    Deep learning for the design of photonic structures.Nature photonics, 15(2):77–90, 2021

    Wei Ma, Zhaocheng Liu, Zhaxylyk A Kudyshev, Alexandra Boltasseva, Wenshan Cai, and Yongmin Liu. Deep learning for the design of photonic structures.Nature photonics, 15(2):77–90, 2021

    2021

  37. [37]

    Piggott, Weiliang Jin, Jelena Vuck- ović, and Alejandro W

    Sean Molesky, Zin Lin, Alexander Y. Piggott, Weiliang Jin, Jelena Vuck- ović, and Alejandro W. Rodriguez. Inverse design in nanophotonics. Nature Photonics, 12(11):659–670, 2018

    2018

  38. [38]

    A portable and stealthy inaudible voice attack based on acoustic metamaterials.IEEE Transactions on Information Forensics and Security, 2025

    Zhiyuan Ning, Juan He, Zhanyong Tang, Weihang Hu, and Xiaojiang Chen. A portable and stealthy inaudible voice attack based on acoustic metamaterials.IEEE Transactions on Information Forensics and Security, 2025

    2025

  39. [39]

    Metaguardian: Enhancing voice assistant security through advanced acoustic meta- materials

    Zhiyuan Ning, Zheng Wang, and Zhanyong Tang. Metaguardian: Enhancing voice assistant security through advanced acoustic meta- materials. InProceedings of the 31st Annual International Conference on Mobile Computing and Networking, pages 788–801, 2025

    2025

  40. [40]

    Topology optimization of acous- tic metasurfaces by using a two-scale homogenization method.Applied Mathematical Modelling, 98:465–497, 2021

    Yuki Noguchi and Takayuki Yamada. Topology optimization of acous- tic metasurfaces by using a two-scale homogenization method.Applied Mathematical Modelling, 98:465–497, 2021

    2021

  41. [41]

    Multilayer perceptron and neural networks.WSEAS transactions on circuits and systems, 8(7):579–588, 2009

    Marius-Constantin Popescu, Valentina E Balas, Liliana Perescu- Popescu, and Nikos Mastorakis. Multilayer perceptron and neural networks.WSEAS transactions on circuits and systems, 8(7):579–588, 2009

    2009

  42. [42]

    Broadband sound absorption using multiple hybrid resonances of acoustic metasurfaces.International Journal of Mechanical Sciences, 229:107508, 2022

    Hyeonbin Ryoo and Wonju Jeon. Broadband sound absorption using multiple hybrid resonances of acoustic metasurfaces.International Journal of Mechanical Sciences, 229:107508, 2022

    2022

  43. [43]

    Learning to modulate pre-trained models in rl.Advances in Neural Information Processing Systems, 36:38231–38265, 2023

    Thomas Schmied, Markus Hofmarcher, Fabian Paischer, Razvan Pas- canu, and Sepp Hochreiter. Learning to modulate pre-trained models in rl.Advances in Neural Information Processing Systems, 36:38231–38265, 2023

    2023

  44. [44]

    DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Yang Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  45. [45]

    Nelder-mead algorithm.Scholarpedia, 4(7):2928, 2009

    Saša Singer and John Nelder. Nelder-mead algorithm.Scholarpedia, 4(7):2928, 2009

    2009

  46. [46]

    Numeracy for language models: Evaluating and improving their ability to predict numbers

    Georgios Spithourakis and Sebastian Riedel. Numeracy for language models: Evaluating and improving their ability to predict numbers. In Proceedings of the 56th Annual Meeting of the Association for Computa- tional Linguistics (Volume 1: Long Papers), pages 2104–2115, 2018

    2018

  47. [47]

    Broad- band acoustic focusing via binary rectangular cavity/helmholtz res- onator metasurface.Journal of Applied Physics, 129(15), 2021

    Shuai Tang, Bin Ren, Yuxin Feng, Jie Song, and Yongyuan Jiang. Broad- band acoustic focusing via binary rectangular cavity/helmholtz res- onator metasurface.Journal of Applied Physics, 129(15), 2021

    2021

  48. [48]

    Attention is all you need.Advances in neural information processing systems, 30, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017

    2017

  49. [49]

    Metasonic: Advancing robot localization with directional embedded acoustic signals.IEEE Robotics and Automation Letters, 2025

    Junling Wang, Zhenlin An, and Yi Guo. Metasonic: Advancing robot localization with directional embedded acoustic signals.IEEE Robotics and Automation Letters, 2025

    2025

  50. [50]

    Wavefront modulation and sub- wavelength diffractive acoustics with an acoustic metasurface.Nature communications, 5(1):5553, 2014

    Yangbo Xie, Wenqi Wang, Huanyang Chen, Adam Konneker, Bogdan- Ioan Popa, and Steven A Cummer. Wavefront modulation and sub- wavelength diffractive acoustics with an acoustic metasurface.Nature communications, 5(1):5553, 2014

    2014

  51. [51]

    Inverse design on customised absorption of acoustic metamaterials with high degrees of freedom by deep learning.Mechanical Systems and Signal Processing, 237:112989, 2025

    Jiahui Yan, Yingli Li, Guohui Yin, Song Yao, Yong Peng, et al. Inverse design on customised absorption of acoustic metamaterials with high degrees of freedom by deep learning.Mechanical Systems and Signal Processing, 237:112989, 2025

    2025

  52. [52]

    Development of adjustable paral- lel helmholtz acoustic metamaterial for broad low-frequency sound absorption band.Materials, 15(17):5938, 2022

    Xiaocui Yang, Fei Yang, Xinmin Shen, Enshuai Wang, Xiaonan Zhang, Cheng Shen, and Wenqiang Peng. Development of adjustable paral- lel helmholtz acoustic metamaterial for broad low-frequency sound absorption band.Materials, 15(17):5938, 2022

    2022

  53. [53]

    Flat optics with designer metasur- faces.Nature materials, 13(2):139–150, 2014

    Nanfang Yu and Federico Capasso. Flat optics with designer metasur- faces.Nature materials, 13(2):139–150, 2014

    2014

  54. [54]

    Learning to inversely design acoustic metamaterials for enhanced performance

    Hongjia Zhang, Jiawei Liu, Weitong Ma, Haitao Yang, Yang Wang, Haibin Yang, Honggang Zhao, Dianlong Yu, and Jihong Wen. Learning to inversely design acoustic metamaterials for enhanced performance. Acta Mechanica Sinica, 39(7):722426, 2023

    2023

  55. [55]

    Remote whispering metamaterial for non-radiative transceiving of ultra-weak sound.Nature Communications, 12(1):3670, 2021

    Jin Zhang, Wei Rui, Chengrong Ma, Ying Cheng, Xiaojun Liu, and Johan Christensen. Remote whispering metamaterial for non-radiative transceiving of ultra-weak sound.Nature Communications, 12(1):3670, 2021

    2021

  56. [56]

    A Survey of Large Language Models

    Wayne Xin Zhao, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, et al. A survey of large language models.arXiv preprint arXiv:2303.18223, 1(2):1–124, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  57. [57]

    Metagen: Llm-driven generative framework for intelligent metasurface element

    Yizhe Zhao, Lili Chen, Xinyi Li, Yue-Jiang Dong, Ju Ren, and Yaoxue Zhang. Metagen: Llm-driven generative framework for intelligent metasurface element. InProceedings of the 23rd Annual International Conference on Mobile Systems, Applications and Services, pages 333–346, 2025

    2025

  58. [58]

    M2silent: Enabling multi-user silent speech interactions via multi-directional speakers in shared spaces

    Juntao Zhou, Dian Ding, Yijie Li, Yu Lu, Yida Wang, Yongzhao Zhang, Yi-Chao Chen, and Guangtao Xue. M2silent: Enabling multi-user silent speech interactions via multi-directional speakers in shared spaces. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pages 1–19, 2025

    2025

  59. [59]

    Reverse design of broadband sound absorption structure based on deep learning method

    Yihong Zhou, Lifeng Ma, Xi Kang, and Zhiyuan Zhu. Reverse design of broadband sound absorption structure based on deep learning method. Physics-Guided Sequence-Based Generative Framework for Acoustic Metamaterial Inverse Design Conference’17, July 2017, Washington, DC, USA Scientific Reports, 15:1946, 2025

    2017

  60. [60]

    Al- gorithm 778: L-bfgs-b: Fortran subroutines for large-scale bound- constrained optimization.ACM Transactions on mathematical software (TOMS), 23(4):550–560, 1997

    Ciyou Zhu, Richard H Byrd, Peihuang Lu, and Jorge Nocedal. Al- gorithm 778: L-bfgs-b: Fortran subroutines for large-scale bound- constrained optimization.ACM Transactions on mathematical software (TOMS), 23(4):550–560, 1997

    1997

  61. [61]

    Parameter prediction and inverse design of metamaterial acoustic absorber based on artificial neural network.Acta Acustica, 50(2):257–267, 2025

    Yuanzhou Zhu, Chun Gong, Ye Gu, Houyou Long, Ying Cheng, and Xiaojun Liu. Parameter prediction and inverse design of metamaterial acoustic absorber based on artificial neural network.Acta Acustica, 50(2):257–267, 2025. A Overlap Checker The overlap checker is a geometric validation algorithm (Algo. 1) designed to verify the spatial feasibility of a struc-...

    2025