Pith. sign in

REVIEW 3 major objections 2 minor 54 references

Proprioception and multi-contact touch let a physics-guided generative model complete objects under heavy hand occlusion at real metric scale.

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-4.5

2026-07-12 23:28 UTC pith:SALUJP6J

load-bearing objection The package is broken: 2604.09100’s claims rest on an abstract only; the supplied full text is an unrelated photonics paper, so the reconstruction results cannot be checked. the 3 major comments →

arxiv 2604.09100 v2 pith:SALUJP6J submitted 2026-04-10 cs.CV cs.RO

Physically Grounded 3D Generative Reconstruction under Hand Occlusion using Proprioception and Multi-Contact Touch

Gabriele Mario Caddeo , Pasquale Marra , Lorenzo Natale This is my paper
classification cs.CV cs.RO
keywords amodal 3D reconstructionhand occlusionproprioceptionmulti-contact touchsigned distance fieldflow-matching diffusionphysics-guided generationmetric-scale pose estimation
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.

Vision-only 3D generators struggle when a hand heavily occludes an object: the hidden surface is under-constrained and scale is ambiguous. This paper claims that adding two physical signals—proprioceptive hand pose and multi-contact touch—plus explicit physics losses can resolve that ambiguity. Object structure is encoded as a pose-aware, camera-aligned signed distance field inside a compact latent space; a conditional flow-matching diffusion model is pretrained on clean images and then finetuned on occluded manipulation scenes while conditioning on RGB, visibility masks, a hand latent, and tactile contacts. Differentiable decoder guidance further penalizes hand–object interpenetration and aligns the surface with observed contacts. The result is a metric, physically consistent structure estimate that plugs into ordinary two-stage reconstruction pipelines and, in simulation and on a real humanoid, produces more complete and correctly scaled shapes than vision-only baselines.

Core claim

Conditioning a flow-matching diffusion model over a pose-aware SDF latent space on proprioceptive hand geometry and multi-contact tactile evidence, together with physics-based decoder guidance, substantially improves amodal completion under severe hand occlusion and yields reconstructions that are both physically plausible and correctly scaled in real-world metric units, outperforming vision-only generative baselines and transferring to a robot whose end-effector was never seen in training.

What carries the argument

A Structure-VAE that compresses a pose-aware, camera-aligned signed distance field, followed by a conditional flow-matching diffusion model finetuned with physics objectives (interpenetration reduction and contact alignment) and differentiable decoder guidance at inference.

Load-bearing premise

That the combination of encoded hand geometry and multi-contact touch, once fed into the conditioning and physics losses, is strong enough in practice to pin down the occluded surface and its absolute scale even when real tactile noise and contact uncertainty are present.

What would settle it

Run the identical model on the same occluded manipulation scenes but ablate all tactile and proprioceptive channels (and the physics guidance); if completion accuracy, interpenetration metrics, and metric-scale error do not degrade markedly relative to the full multimodal model, the central claim fails.

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

If this is right

  • Existing two-stage 3D pipelines can replace their first-stage structure estimator with this metric, physically consistent SDF without redesigning the appearance module.
  • Robots that already sense joint angles and multi-contact touch can obtain denser object models during ordinary grasping without extra cameras or depth sensors.
  • Training can begin on large vision-only corpora and still specialize to occluded manipulation with comparatively modest finetuning data.
  • The same conditioning pattern generalizes across end-effector morphologies, reducing the need to retrain for every new gripper or hand.

Where Pith is reading between the lines

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

  • If contact alignment is the dominant scale cue, sparse or noisy tactile arrays may still suffice provided the physics guidance remains active, suggesting cheaper sensors could be viable.
  • The same hand-latent-plus-contact recipe could be applied to other generative 3D representations (occupancy networks, Gaussian splats) that currently ignore physical interaction.
  • Failure modes under soft or deformable objects would be diagnostic: if interpenetration penalties dominate, the method may over-smooth compliant surfaces.
  • Cascading the model across sequential grasps could accumulate a more complete object model without requiring a single fully observed view.

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

3 major / 2 minor

Summary. The submission (arXiv:2604.09100) claims a multimodal, physically grounded pipeline for metric-scale amodal object reconstruction and pose estimation under severe hand occlusion. Object structure is represented as a pose-aware, camera-aligned SDF compressed by a Structure-VAE; a conditional flow-matching diffusion model is pretrained on vision-only data and finetuned on occluded manipulation scenes, conditioned on visible RGB, visibility/occluder masks, a proprioceptive hand latent, and multi-contact tactile signals. Physics-based objectives and differentiable decoder-guidance are used at finetuning and inference to penalize hand–object interpenetration and to align the surface with contact observations. The abstract asserts substantial gains over vision-only baselines in simulation (completion under occlusion, physical plausibility, correct real-world scale) and transfer to a real humanoid whose end-effector differs from those seen in training, with the metric structure estimate intended to plug into existing two-stage geometry/appearance pipelines.

Significance. If the claimed gains hold under rigorous ablations and real contact noise, the work would be a meaningful step for robotic manipulation and amodal 3D reconstruction: proprioception plus multi-contact touch as generative constraints is a natural and under-used inductive bias, and a metric, physically consistent structure estimate that integrates into two-stage pipelines would be practically useful. The abstract also advertises cross-embodiment transfer, which—if demonstrated with clear metrics—would strengthen the contribution beyond simulation-only occlusion completion. Those strengths cannot currently be credited as verified results because the supplied full manuscript does not contain the corresponding methods, experiments, or numbers.

major comments (3)
  1. Manuscript identity mismatch: the review package labels paper_id 2604.09100 (physically grounded 3D amodal reconstruction under hand occlusion; cs.CV), but the FULL MANUSCRIPT TEXT is the complete unrelated paper on ultrafast all-optical switching via a supersolid phase transition of light (driven-dissipative nonlocal GP equation, Lindhard kernel, S-curve bistability, Bragg extinction ratios, multi-layer reconfigurability). No Structure-VAE, pose-aware camera-aligned SDF, flow-matching diffusion, hand latent, tactile conditioning, interpenetration/contact objectives, decoder-guidance, simulation ablations, metric-scale tables, or real-humanoid transfer results appear. The central empirical claim therefore has no inspectable support beyond the abstract.
  2. Because the correct full text is absent, load-bearing elements of the strongest claim cannot be checked: (i) formulation and weighting of physics objectives / decoder-guidance for interpenetration and contact alignment; (ii) how multi-contact touch and proprioceptive hand geometry are encoded and conditioned in the latent flow-matching model; (iii) quantitative ablations isolating proprioception vs. touch vs. vision-only; (iv) metric-scale and physical-plausibility metrics and baselines; (v) real-robot protocol and end-effector transfer evidence. Without those sections, the review cannot assess soundness of the weakest assumption—that contact and interpenetration terms sufficiently resolve occluded metric SDF ambiguity under real noise.
  3. Recommendation cannot be made on scientific merit of 2604.09100 until the authors (or editorial office) supply the matching full manuscript. A resubmission with the correct PDF/source is required before any accept/revise decision on the CV claims is possible.
minor comments (2)
  1. Abstract alone is clear on high-level pipeline (Structure-VAE → conditional flow-matching; pretrain then finetune; physics decoder-guidance) but cannot substitute for missing method equations, architecture details, loss definitions, or experimental tables.
  2. The provided supersolid-switching manuscript (wrong paper) is internally coherent as a physics Letter with SM, but that is irrelevant to refereeing 2604.09100 and should not be scored as the submission under review.

Circularity Check

0 steps flagged

No significant circularity: the supplied full text is a self-contained driven-dissipative GP / Lindhard simulation study; claimed 3D-reconstruction paper is not present to inspect.

full rationale

The package labels paper_id 2604.09100 (physically grounded amodal SDF reconstruction under hand occlusion) but the only full manuscript provided is the unrelated photonic supersolid-switch Letter (driven-dissipative nonlocal GP, drifted-2DEG Lindhard kernel, S-curve bistability, write–hold–erase protocol). For that supplied text, the derivation chain is ordinary open-system mean-field physics: the GP equation and Lindhard polarizability are stated, the uniform-state cubic and Bogoliubov roton growth rate are re-derived in the SM, and the 124 dB contrast and multiport protocols are numerical outcomes of those equations under an explicit pump schedule—not quantities forced by fitting a target or by renaming an input. The sole self-citation ([21], same authors) supplies the prior demonstration that a drifted 2DEG can produce a negative finite-k kernel; that is sequential platform-building, not a load-bearing uniqueness theorem or a prediction that equals its fit by construction. No self-definitional loop, no fitted-input-called-prediction, and no ansatz smuggled as a theorem appear. For the abstract-only 3D-reconstruction claim, the work is an empirical systems/ML pipeline (Structure-VAE + conditional flow-matching + physics decoder-guidance); such pipelines do not reduce their reported occlusion gains to a definitional identity. Residual risk that contact physics used in training is the same physics scored at test time is ordinary for physics-informed models and does not meet the bar for circularity. Score 0; steps empty.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 1 invented entities

Abstract-only review of a robotics/CV systems paper. Load-bearing premises are modeling choices and experimental assumptions rather than mathematical axioms. Free parameters (loss weights, latent dims, guidance scales, contact thresholds) are implied but not quantified in the abstract. No new physical particles or forces are invented; 'Structure-VAE' and physics-guided decoder conditioning are methodological constructs.

free parameters (3)
  • Physics objective / decoder-guidance weights (interpenetration and contact alignment)
    Abstract states physics-based objectives and differentiable decoder-guidance are used at finetuning and inference; relative weights and schedules are free design choices that directly affect the claimed physical consistency.
  • Structure-VAE latent dimensionality and conditioning channels
    Compact latent space and conditioning on RGB, masks, hand latent, and tactile info are architectural free choices not fixed by theory in the abstract.
  • Pretrain/finetune data mixture and occlusion distribution
    Performance under 'severe hand occlusion' depends on how simulated contacts, sensors, and occlusions are generated; those distributions act as free experimental parameters.
axioms (4)
  • domain assumption Proprioceptive posed hand geometry plus multi-contact touch sufficiently constrain occluded object surfaces for metric SDF completion beyond vision-only generative priors.
    Core premise of the abstract: physical interaction signals reduce ambiguity in occluded regions.
  • domain assumption A pose-aware camera-aligned SDF latent space learned by a Structure-VAE is an adequate structure representation for amodal reconstruction and downstream two-stage refinement.
    Representation choice on which the diffusion model and metric claims rest.
  • domain assumption Simulation contact/tactile models and training end-effectors transfer sufficiently to a real humanoid with a different end-effector.
    Abstract claims real-robot validation with a different end-effector; sim-to-real and cross-embodiment transfer is assumed to hold.
  • domain assumption Conditional flow-matching in the Structure-VAE latent space with the listed conditioners is a valid generative model for the occluded structure posterior.
    Standard generative-modeling assumption for the method class.
invented entities (1)
  • Structure-VAE (pose-aware camera-aligned SDF latent space) no independent evidence
    purpose: Compress object structure into a latent code for conditional flow-matching diffusion under occlusion.
    Named methodological construct in the abstract; may be a paper-specific architecture rather than a new physical entity. No independent evidence outside this work is given in the abstract.

pith-pipeline@v1.1.0-grok45 · 24299 in / 3205 out tokens · 30638 ms · 2026-07-12T23:28:02.838445+00:00 · methodology

0 comments
read the original abstract

We propose a multimodal, physically grounded approach for metric-scale amodal object reconstruction and pose estimation under severe hand occlusion. Unlike prior occlusion-aware 3D generation methods that rely only on vision, we leverage physical interaction signals: proprioception provides the posed hand geometry, and multi-contact touch constrains where the object surface must lie, reducing ambiguity in occluded regions. We represent object structure as a pose-aware, camera-aligned signed distance field (SDF) and learn a compact latent space with a Structure-VAE. In this latent space, we train a conditional flow-matching diffusion model, pretraining on vision-only images and finetuning on occluded manipulation scenes while conditioning on visible RGB evidence, occluder/visibility masks, the hand latent representation, and tactile information. Crucially, we incorporate physics-based objectives and differentiable decoder-guidance during finetuning and inference to reduce hand--object interpenetration and to align the reconstructed surface with contact observations. Because our method produces a metric, physically consistent structure estimate, it integrates naturally into existing two-stage reconstruction pipelines, where a downstream module refines geometry and predicts appearance. Experiments in simulation show that adding proprioception and touch substantially improves completion under occlusion and yields physically plausible reconstructions at correct real-world scale compared to vision-only baselines; we further validate transfer by deploying the model on a real humanoid robot with an end-effector different from those used during training.

Figures

Figures reproduced from arXiv: 2604.09100 by Gabriele Mario Caddeo, Lorenzo Natale, Pasquale Marra.

Figure 1
Figure 1. Figure 1: Inference pipeline. We present a method to generate physically plausible 3D shape reconstructions by fusing vision with contact. In particular, we consider the active information of contact coming from tactile sensors distributed on the hand, and the negative information (non-interpenetration) coming from the hand geometry. Apart from these information, the method requires the Egocentric RGB Image of the o… view at source ↗
Figure 2
Figure 2. Figure 2: Training procedure and Architecture. a1): We train a Structure-VAE autoencoder that reconstructs pose-aware object SDFs. a2): Using the frozen VAE en￾coder, we build latent datasets and train a conditional flow transformer from scratch using pose-consistent, unoccluded object images. a3): We finetune the Structure￾flow on occluded manipulation scenes, conditioning on visible RGB evidence, oc￾cluder/visibil… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative results in simulation. Vision-only baselines often exhibit ar￾tifacts (e.g., holes or inconsistent relative dimensions) under occlusion. By leveraging contact cues, our method produces more physically plausible reconstructions. proprioception, touch, and physics-based constraints improves performance on all metrics, except in the least-occluded regime for F@0.02. This suggests that when occlusi… view at source ↗
Figure 4
Figure 4. Figure 4: 4.4 Ablation Study We perform ablations to quantify the contribution of individual components of our method and to evaluate robustness to noise in tactile readings. Additional [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative results on real-world data. Contact cues improve physical plausibility under occlusion. Artifacts can arise when camera–hand calibration/forward kinematics are inaccurate, leading to hand–object misalignment (third row). qualitative examples and extended quantitative results are provided in the sup￾plementary material. Sensing ablation. To assess the role of each sensing modality, we train vari… view at source ↗

Review history (2 revisions) →

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

54 extracted references

  1. [1]

    D. A. B. Miller, Nature Photon.4, 3 (2010)

    2010

  2. [2]

    D. A. B. Miller, Journal of Lightwave Technology35, 346 (2017)

    2017

  3. [3]

    Y. Shen, N. C. Harris, S. Skirlo, M. Prabhu, T. Baehr- Jones, M. Hochberg, X. Sun, S. Zhao, H. Larochelle, D. Englund, and M. Soljačić, Nature Photonics11, 441 (2017)

    2017

  4. [4]

    Nozaki, T

    K. Nozaki, T. Tanabe, A. Shinya, S. Matsuo, T. Sato, H. Taniyama, and M. Notomi, Nature Photon.4, 477 (2010)

    2010

  5. [5]

    H. J. Caulfield and S. Dolev, Nature Photon.4, 261 (2010)

    2010

  6. [6]

    V. R. Almeida, C. A. Barrios, R. R. Panepucci, and M. Lipson, Nature431, 1081 (2004)

    2004

  7. [7]

    Tanabe, M

    T. Tanabe, M. Notomi, S. Mitsugi, A. Shinya, and E. Kuramochi, Appl. Phys. Lett.87, 151112 (2005)

    2005

  8. [8]

    Wuttig and N

    M. Wuttig and N. Yamada, Nature Mater.6, 824 (2007)

    2007

  9. [9]

    C. Rios, M. Stegmaier, P. Hosseini, D. Wang, T. Scherer, C. D. Wright, H. Bhaskaran, and W. H. P. Pernice, Nature Photon.9, 725 (2015)

    2015

  10. [10]

    Lett.15, 1479 (2003)

    J.Mørk, F.Ohman, andS.Bischoff,IEEEPhoton.Tech- nol. Lett.15, 1479 (2003)

    2003

  11. [11]

    Ballarini, M

    D. Ballarini, M. De Giorgi, E. Cancellieri, R. Houdré, E. Giacobino, R. Cingolani, A. Bramati, G. Gigli, and D. Sanvitto, Nat. Commun.4, 1778 (2013)

    2013

  12. [12]

    A. Amo, J. Lefrère, S. Pigeon, C. Adrados, C. Ciuti, I. Carusotto, R. Houdré, E. Giacobino, and A. Bramati, Nature Phys.5, 805 (2009)

    2009

  13. [13]

    Sanvitto and S

    D. Sanvitto and S. Kéna-Cohen, Nature Mater.15, 1061 (2016)

    2016

  14. [14]

    Carusotto and C

    I. Carusotto and C. Ciuti, Rev. Mod. Phys.85, 299 (2013)

    2013

  15. [15]

    K. G. Lagoudakis, M. Wouters, M. Richard, A. Baas, I. Carusotto, R. André, L. S. Dang, and B. Deveaud- Plédran, Nature Phys.4, 706 (2008)

    2008

  16. [16]

    Klaers, J

    J. Klaers, J. Schmitt, F. Vewinger, and M. Weitz, Nature 468, 545 (2010)

    2010

  17. [17]

    R. C. Schofield, P. R. Dolan, J. M. Smith,et al., Nature 6 Photon.18, 1083 (2024)

    2024

  18. [19]

    Weisbuch, M

    C. Weisbuch, M. Nishioka, A. Ishikawa, and Y. Arakawa, Phys. Rev. Lett.69, 3314 (1992)

    1992

  19. [20]

    Kasprzak, M

    J. Kasprzak, M. Richard, S. Kundermann, A. Baas, P. Jeambrun, J. M. J. Keeling, F. M. Marchetti, M. H. Szymańska, R. André, J. L. Staehli, V. Savona, P. B. Lit- tlewood, B. Deveaud, and L. S. Dang, Nature443, 409 (2006)

    2006

  20. [21]

    J. L. Figueiredo, J. T. Mendonça, and H. Terças, Phys. Rev. A113, L031303 (2026)

    2026

  21. [22]

    Tanzi, E

    L. Tanzi, E. Lucioni, F. Famà, J. Catani, A. Fioretti, C. Gabbanini, R. N. Bisset, L. Santos, and G. Modugno, Phys. Rev. Lett.122, 130405 (2019)

    2019

  22. [23]

    M. Guo, F. Böttcher, J. Hertkorn, J.-N. Schmidt, M.Wenzel, H.P.Büchler, T.Pfau, andI.Ferrier-Barbut, Nature574, 386 (2019)

    2019

  23. [24]

    M. A. Norcia, C. Politi, L. Klaus, E. Poli, M. Sohmen, M. J. Mark, R. N. Bisset, L. Santos, and F. Ferlaino, Nature596, 357 (2021)

    2021

  24. [25]

    Böttcher, J.-N

    F. Böttcher, J.-N. Schmidt, M. Wenzel, J. Hertkorn, M. Guo, T. Langen, and T. Pfau, Phys. Rev. X9, 011051 (2019)

    2019

  25. [26]

    Léonard, A

    J. Léonard, A. Morales, P. Zupancic, T. Esslinger, and T. Donner, Nature543, 87 (2017)

    2017

  26. [27]

    J.-R. Li, J. Lee, W. Huang, S. Burchesky, B. Shteynas, F. C. Top, A. O. Jamison, and W. Ketterle, Nature543, 91 (2017)

    2017

  27. [28]

    See Supplemental Material at [URL] for details on the Gross-Pitaevskii equation, experimental scheme, addi- tional simulation results and characterizations of the switching protocol, and proposed generalization for higher number of states

  28. [29]

    L. P. Pitaevskii and S. Stringari,Bose–Einstein Conden- sation(Oxford University Press, Oxford, 2003)

    2003

  29. [30]

    Dalfovo, S

    F. Dalfovo, S. Giorgini, L. P. Pitaevskii, and S. Stringari, Rev. Mod. Phys.71, 463 (1999)

    1999

  30. [31]

    Wouters and I

    M. Wouters and I. Carusotto, Phys. Rev. Lett.99, 140402 (2007)

    2007

  31. [32]

    P. D. Drummond and D. F. Walls, J. Phys. A: Math. Gen.13, 725 (1980)

    1980

  32. [33]

    H. M. Gibbs,Optical Bistability: Controlling Light with Light(Academic Press, Orlando, 1985)

    1985

  33. [34]

    H. Ju, S. Zhang, D. Lenstra, H. de Waardt, E. Tang- diongga, G. D. Khoe, and H. J. Dorren, Opt. Express 13, 942 (2005)

    2005

  34. [35]

    D. E. Chang, V. Vuletić, and M. D. Lukin, Nature Pho- ton.8, 685 (2014)

    2014

  35. [36]

    R. S. Fishman and S. H. Liu, Phys. Rev. Lett.76, 2398 (1996)

    1996

  36. [37]

    Kaplan, P

    D. Kaplan, P. A. Volkov, A. Cavalleri, and P. Chandra, arXiv preprint arXiv:2511.07320 (2025)

    arXiv 2025

  37. [38]

    Gao and E

    Q. Gao and E. Khalaf, Phys. Rev. B106, 075420 (2022)

    2022

  38. [39]

    Gossard, and W

    H.L.Störmer, Z.Schlesinger, A.Chang, D.C.Tsui, A.C. Gossard, and W. Wiegmann, Phys. Rev. Lett.51, 126 (1983)

    1983

  39. [40]

    Pfeiffer, K

    L. Pfeiffer, K. W. West, H. L. Stormer, and K. W. Bald- win, Appl. Phys. Lett.55, 1888 (1989)

    1989

  40. [41]

    G. T. Reed, G. Mashanovich, F. Y. Gardes, and D. J. Thomson, Nature Photon.4, 518 (2010)

    2010

  41. [42]

    D. N. Basov, M. M. Fogler, and F. J. García de Abajo, Science354, aag1992 (2016)

    2016

  42. [43]

    T. Low, A. Chaves, J. D. Caldwell, A. Kumar, N. X. Fang, P. Avouris, T. F. Heinz, F. Guinea, L. Martin- Moreno, and F. Koppens, Nature Mater.16, 182 (2017)

    2017

  43. [44]

    Ostermann, F

    S. Ostermann, F. Piazza, and H. Ritsch, Phys. Rev. X 6, 021026 (2016)

    2016

  44. [45]

    M. J. Hartmann, F. G. S. L. Brandão, and M. B. Plenio, Laser Photon. Rev.2, 527 (2008)

    2008

  45. [46]

    Ultrafast All-Optical Switching via a Supersolid Phase Transition of Light

    A. Tomadin and R. Fazio, J. Opt. Soc. Am. B27, A130 (2010). Supplemental Material: “Ultrafast All-Optical Switching via a Supersolid Phase Transition of Light” J. L. Figueiredo, 1 J. T. Mendon¸ ca,1 and H. Ter¸ cas2, 1 1GoLP/Instituto de Plasmas e Fus˜ ao Nuclear, Instituto Superior T´ ecnico, Universidade de Lisboa, 1049-001 Lisboa, Portugal 2Instituto S...

    2010

  46. [47]

    For Γ hold = 0.5 the memory is not retained:N(t) decays to the OFF branch after the write pulse. The reason is dynamical rather than static–the upper branch at Γ = 0.5 is a valid steady-state solution (N ON = 2.64), but the large transient overshoot (N≃7) generated by the write pulse drives the field so far above the upper branch that the subsequent relax...

  47. [48]

    J. L. Figueiredo, J. T. Mendon¸ ca, and H. Ter¸ cas, Phys. Rev. A113, L031303 (2026)

    2026

  48. [49]

    Kruchkov, Phys

    A. Kruchkov, Phys. Rev. A89, 033862 (2014)

    2014

  49. [50]

    R. C. Schofield, M. Fu, E. Clarke, I. Farrer, A. Trapalis, H. S. Dhar, R. Mukherjee, T. Severs Millard, J. Heffer- nan, F. Mintert, R. A. Nyman, and R. F. Oulton, Nature Photonics18, 1083 (2024)

    2024

  50. [51]

    Pieczarkaet al., Nature Photon.18, 1090 (2024)

    M. Pieczarkaet al., Nature Photon.18, 1090 (2024)

    2024

  51. [52]

    W. Bao, D. Jaksch, and P. A. Markowich, Journal of Computational Physics187, 318 (2003)

    2003

  52. [53]

    J. A. C. Weideman and B. M. Herbst, SIAM Journal on Numerical Analysis23, 485 (1986), https://doi.org/10.1137/0723033

    doi:10.1137/0723033 1986

  53. [54]

    H. L. Stormer, Z. Schlesinger, A. Chang, D. C. Tsui, A. C. Gossard, and W. Wiegmann, Phys. Rev. Lett.51, 126 (1983)

    1983

  54. [55]

    Pfeiffer, K

    L. Pfeiffer, K. West, H. Stormer, and K. Baldwin, Applied Physics Letters55, 1888 (1989). 8 � � ����� ��� ���������� �������� ���� � �� � � ��� � �� � � ��� � �� �� � ��� ����� ��� ��� ��� ��� ��� � ��� �� � �� ��� � ��� ��� ��� ��� ��� � ��� � ��� ��� �� � �� ��� � ��� ��� ��� ��� ��� ��� ��� ��� � ���� ��� � ��� � ��� � ��� � ��� ����� ��� ��� ��� �����...

    1989