Pith. sign in

REVIEW 4 major objections 4 minor 47 references

FillGauss: Fine-Grained Filling-Aware Impact Sound Generation for 3D Gaussian Splatting

T0 review · 4 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Containers sound different when filled, and a 3D-aware diffusion model can synthesize that difference.

desk verdict Useful new dataset and a plausible fill-aware conditioning scheme, but the evaluation overreaches: the split may leak object identity and the SOTA claim trips over the paper's own FAD table. read the letter →

arxiv 2607.17773 v1 pith:23JEBDFP submitted 2026-07-20 cs.MM

Chen Yang , Ganye Wen , Bin Huang , Jiayi Lyu , Zehai Niu , Linlin Shen , Jinbao Wang This is my paper · ORCID
classification cs.MM
keywords Impactsoundgeneration3DGaussianSplattinglatentdiffusionfilling-awareaudiocross-modalsynthesisFilldatasetacousticresonanceaudio-visual
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that the sound of a struck container depends on its internal fill state, not just its outer geometry, and that existing 3D-aware audio generators fail because they treat objects as hollow shells. To fix this, it introduces FillImpact, a dataset of over 5,000 real impact recordings from 88 containers with controlled fill materials (water and rice), fill levels, and striker materials, plus 3D Gaussian Splatting scans and exact 3D strike coordinates. It then proposes FillGauss, a latent-diffusion framework that fuses 3D geometry, strike position, and a textual description of striker and fill state. Experiments report that FillGauss produces audio whose dominant frequency drops with fill level, whose damping differs between water and rice, and whose brightness differs by striker hardness, claiming state-of-the-art physical grounding for cross-modal audio generation. A sympathetic reader should care because this is a step toward interactive and robotic settings where the internal state of a container is visually hidden but acoustically audible.

What carries the argument

The load-bearing mechanism is the cross-modal attention fusion pair: context self-attention combines the positional embedding of the strike point with the text embedding of the physical state into a unified excitation context; then geometry cross-attention lets the 3D Gaussian features query that context, so each surface region selectively absorbs positional and damping semantics. This fused embedding conditions a diffusion-transformer denoiser, trained with Huber loss, that predicts the audio latent. The dataset's verified physical trends (dominant frequency decreasing with fill level, higher decay for rice than water, higher spectral centroid for steel) provide the prior that the model is

What would settle it

An object-disjoint evaluation: hold out all recordings of a set of containers and measure FAD on impacts of those unseen containers. If FillGauss's FAD degrades substantially compared to the random-split result, the claimed physical generalization is not established. A second check is to synthesize audio for the same 3D scan with two different fill descriptions and verify the dominant-frequency shift quantitatively.

Watch

Extended reading notes

Core claim

On its own terms, the central discovery is that internal filling state is a first-class acoustic condition: adding mass lowers resonance frequency according to the harmonic-oscillator relation f ∝ sqrt(k/m), granular fill suppresses resonance through high damping, and harder strikers shift energy to higher frequencies. FillGauss represents an object as 3D Gaussians, encodes the exact strike point with positional encoding, encodes fill and striker as natural-language text, and uses a self-attention-plus-cross-attention fusion to make the 3D surface query the localized physical context. The fused condition guides a pretrained latent audio diffusion model. The paper shows qualitative spectrogra

Load-bearing premise

The reported 80/20 train/test split is described only over samples, not over objects; if the same container appears in both training and testing at different fill levels, the model could memorize object-specific resonances rather than learn transferable physics.

Editorial extensions

If this is right

  • If FillGauss works as claimed, interactive 3D scenes can add plausible impact sounds for containers whose fill state changes, without re-recording audio for every state.
  • The dependence on a textual fill description means fill state can be controlled at inference time: the same 3D scan can sound hollow, water-filled, or rice-filled on demand.
  • Because the model is conditioned on exact 3D strike coordinates, sounds can vary continuously with strike position across the surface, enabling positional audio feedback in virtual reality and robotic manipulation.
  • The finding that fine-grained text prompts outperform hard-coded categorical embeddings suggests pretrained language models are an effective carrier for physical condition semantics.
  • The sparsity experiment indicates that a compressed 3D Gaussian representation still yields acceptable audio, pointing to a practical route for real-time synthesis.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If fill state is truly acoustically discriminable, then the audio itself could be used to infer the fill state of an unseen container, a task the paper does not attempt but its dataset enables.
  • The paper's argument implies a failure mode for purely geometric 3D-aware models: they will average over fill states and produce a generic container sound; one could test this by comparing a hollow-only model on filled containers.
  • The five-level discretization of fill could be extended to continuous regression or to dynamic processes like sloshing, which the paper lists as future work.
  • A practical extension would be to train on object-disjoint splits to verify generalization beyond the 88 recorded containers.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper introduces a new task, fine-grained filling-aware impact sound generation, and contributes FillImpact, a dataset of over 5,000 recordings from 88 real containers with varied fill materials, fill levels, strikers, and strike positions, together with 3DGS reconstructions. It then proposes FillGauss, a latent-diffusion framework that fuses 3DGS geometric features, a 3D strike coordinate, and text-described physical state through self- and cross-attention to condition audio generation on the internal fill state. The paper reports physical acoustic analysis of the dataset, quantitative comparisons against SonicGauss and TangoFlux, subjective listening tests, ablations, and sparsity sensitivity experiments, claiming a new state-of-the-art for physically grounded cross-modal audio generation. The central novelty is conditioning on internal fill state rather than only surface geometry, and this is a promising direction.

Significance. If the empirical claims are sustained, the work would be a valuable contribution: it defines a concrete task, provides a controlled multimodal dataset with physical annotations, and proposes a principled architecture that combines 3DGS geometry with internal-state conditioning. The physical analysis of the dataset (e.g., frequency shift with fill level, distinct decay for granular vs. liquid fill, striker hardness effects) is a useful prior for future work, and the ablations show the contribution of each component. However, the central generalization claim to unseen 3D-scanned containers is currently unsupported because the train/test split is not stated to be object-disjoint. In addition, the headline state-of-the-art claim conflicts with the paper's own primary metric, and the subjective evaluation has internal inconsistencies. With a corrected evaluation protocol and appropriately qualified claims, the contribution would be solid and publishable.

major comments (4)
  1. [Section V.A (Implementation Details)] The 80/20 split is described only as a 'partition of FillImpact Dataset into an 80% training split and a 20% testing split,' with no statement that the 88 physical objects are split disjointly. Because each object appears in multiple recordings across fill levels, fill materials, strikers, and strike positions, a sample-level split will likely place the same object in both training and test sets. The 3DGS geometry stream in the cross-attention fusion (Eq. 4) can then act as a near-unique object fingerprint, letting the model retrieve object-specific resonances rather than predict them from geometry plus fill state. This would inflate every metric in Tables I–IV and invalidates the claim of generalization to unseen containers. Please report an object-disjoint split with per-object counts, and/or add a held-out-object evaluation.
  2. [Table I vs. Abstract] FAD is declared the primary objective metric (Section V.A), but TangoFlux FT achieves a better FAD (1.0627) than FillGauss (1.3500), and FillGauss is also worse than SonicGauss FT on KL sig (0.6977 vs. 0.5750). The unqualified 'state-of-the-art' claims in the Abstract and Section V.B are therefore not supported by the paper's own primary metric. Either restrict the claim to the best 3D-aware method, or provide a principled argument for why a worse FAD is acceptable (e.g., TangoFlux's lack of spatial grounding) and support it with a metric that privileges 3D conditioning.
  3. [Section V.A / Appendix C / Table II] There is an inconsistency in the reported number of participants: Section V.A states 8 participants, while Appendix C states 20. Additionally, Table II shows FillGauss MOS (4.43) slightly below SonicGauss ZS (4.50), yet FillGauss has a 69.6% win rate; this is not explained. Because the subjective results are central to the physical-grounding claims, the participant count must be corrected and the MOS/win-rate relationship interpreted. Statistical significance testing for the subjective comparisons should also be reported.
  4. [Abstract / Section III / Section IV.A] The task is described as requiring a 'continuous fill level' in the Abstract and IV.A, but Section III states the fill level is discretized into five states (level 0 = hollow). 'Fine-grained' conditioning and interpolation across arbitrary fill levels is not demonstrated with five discrete levels. Clarify the actual conditioning variable and, if continuous interpolation is claimed, add an experiment on unseen intermediate fill levels.
minor comments (4)
  1. [Table III] Table III lists only four attributes (Striker Mat., Pos., Fill Mat., Level), but the text says five attributes including container material are evaluated. Add the missing column and report its matching rate.
  2. [Section V.A] The metric 'KL sig' is not defined; specify how the KL divergence is computed and what 'sig' denotes. Also, FAD is described as 'reference-free,' which is misleading; FAD compares generated and reference feature distributions via VGGish.
  3. [Reproducibility] No code, checkpoints, or split metadata are released. If the object-disjoint split is adopted, release the object IDs and random seed for reproducibility.
  4. [Figure 4 / Section III] The exact normalization used for 'normalized dominant frequency' is not stated. Clarify whether it is per-object or per-level normalization, so the claimed physical trend can be reproduced.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's empirical predictions are not equivalent by construction to their conditioning inputs.

full rationale

FillGauss is an empirical cross-modal generative model, not a derivation of physical laws from first principles. The physical principles cited (f∝sqrt(k/m), damping differences) are introduced from external acoustics literature (e.g., [24]) and used as post-hoc validation of the dataset and of generated spectrograms; they are not outputs derived from the model equations, and the model is not fitted to those laws. Fill-awareness is architecturally induced by conditioning on fill-level text and/or categorical embeddings, but the paper's quantitative and perceptual evaluations test whether the learned model actually uses that conditioning (e.g., Table III attribute matching and Table V sparsity sensitivity), so success is not guaranteed by construction. The candidate concern that the 80/20 split is described only as a sample partition (Section V.A) is a real evaluation-validity risk (possible object-level leakage), but it is not a circularity of the derivation: it concerns whether the test metrics measure generalization, not whether a prediction reduces to an input by definition or via self-citation. No load-bearing self-citation was found: the cited SonicGauss, TangoFlux, SplatFormer, and Stable Audio Open are external prior works, and no uniqueness theorem or ansatz is imported from the present authors' own prior work. Therefore no circular step can be exhibited, and the appropriate score is 0.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper introduces a dataset and a learned generator; it does not introduce new physical entities. The main hidden burden is the sufficiency of 3DGS-plus-text as a conditioning set, and the assumption that the instance-level split supports generalization. No machine-checked proof or physical derivation is offered, so the empirical claims rest on dataset quality and evaluation design.

free parameters (3)
  • Fill-level discretization = 5 levels (0-4)
    The paper's 'continuous fill level' claim is operationalized as five discrete states; the mapping from physical fill fraction to level index is not specified. Central to the fine-grained claim.
  • Huber loss threshold δ = 1.0
    Hand-set in Eq. (6) for diffusion training stability; not central to the physical claim but a design choice.
  • 3DGS sparsity voxel sizes = {0.0, 0.02, 0.05, 0.08, 0.12}
    Used in the sensitivity analysis to define sparsity levels L1-L5; the choice affects the claimed L3 optimal boundary.
assumptions (5)
  • domain assumption Frozen pre-trained Audio VAE and text encoder preserve fine spectral detail needed for fill discrimination.
    Section IV.D / Implementation Details: Audio VAE and Text Encoder are frozen; if their latents blur damping cues, fill-aware synthesis would fail.
  • domain assumption 3DGS surface geometry plus text condition is sufficient to determine impact sound; wall thickness, internal cavity shape, and wall coupling are not modeled.
    Section IV.A problem definition omits these variables; the model has no channel for them.
  • domain assumption The 80/20 split is instance-level, not object-disjoint.
    Section V.A: 'partition FillImpact Dataset into an 80% training split and a 20% testing split'—no statement about object identity separation.
  • domain assumption The harmonic oscillator relation f ∝ sqrt(k/m) applies to filled containers with added mass.
    Section III Acoustic Analysis uses this law to validate fill-level trends; no direct measurement of k or m is reported.
  • domain assumption Granular fill (rice) decays faster than liquid fill (water) and this is a stable ground-truth cue.
    Section III, Figure 4 middle; used both to validate the dataset and to interpret generated spectrograms.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FillGauss: Fine-Grained Filling-Aware Impact Sound Generation for 3D Gaussian Splatting." pith.science (2026). https://pith.science/paper/23JEBDFP

@misc{pith2026260717773,
  author       = {Pith},
  title        = {Pith review of: FillGauss: Fine-Grained Filling-Aware Impact Sound Generation for 3D Gaussian Splatting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/23JEBDFP}},
  note         = {Machine review of arXiv:2607.17773}
}
read the original abstract

Synthesizing physically plausible impact sounds from visual observations remains a great challenge in multi-modal AI. Existing 3D-aware audio generation methods primarily model the surface geometry of hollow rigid bodies. However, they fundamentally overlook internal filling states, a critical physical factor that drastically modulates acoustic resonance and damping. To address this issue, we have defined a new task called Fine-Grained Filling-Aware Impact Sound Generation. As a foundational step, we first introduce the fine-grained fill-aware dataset (FillImpact), a pioneering multi-modal collection comprising over 5,000 rigorous acoustic recordings from 88 diverse real-world objects. It captures impact interactions with varying internal contents (i.e., water, rice), a continuous range of fill levels, and distinct striker materials. Furthermore, comprehensive acoustic analysis confirms that the collected data closely aligns with established physical laws governing acoustic resonance and damping, indicating its suitability for physically grounded modeling. Building on this dataset, we propose a novel generative framework (FillGauss) that integrates 3D Gaussian Splatting (3DGS) with internal state conditioning for sound generation. By fusing 3DGS geometric features, precise 3D spatial strike coordinates, and fine-grained textual physical conditions within a latent diffusion architecture, FillGauss enables position-aware, striker-aware, and filling-aware audio generation. Extensive experiments demonstrate that our approach could generate high-fidelity impact sounds that adhere to underlying physical principles, establishing a new state-of-the-art for physically grounded cross-modal audio generation.

Figures

Figures reproduced from arXiv: 2607.17773 by the authors.

Figure 1
Figure 1. Overview of the FillGauss framework capabilities. Left: Some objects of FillImpact. Right: Given a 3D Gaussian Splatting representation, our model [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Simplified architecture of the FillGauss framework. It illustrates how [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of FillImpact. Left: The construction of FillImpact. Right: Hierarchical distribution of FillImpact. The sunburst chart details the compositional [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Acoustic analysis of FillImpact. Left: The normalized dominant frequency decreases as the fill level increases, strictly following the physical principle [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Overview of FillGauss. The framework first extracts individual features from the 3DGS representation [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Results of position-aware sound generation. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Results of striker-aware sound generation across different tool [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Results of fill-material-aware sound generation. [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: Qualitative results of fill-level-aware sound generation across continuous filling states. [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: Visual comparison of 3DGS reconstructions under varying sparsity [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 13
Figure 13. Figure 13: Stacked bar chart illustrating the combinatorial distribution of [PITH_FULL_IMAGE:figures/full_fig_p011_13.png]
Figure 14
Figure 14. Figure 14: Distribution of physical scales across all collected objects, demon [PITH_FULL_IMAGE:figures/full_fig_p011_14.png]
Figure 12
Figure 12. Figure 12: Proportional distribution of audio recordings based on the internal [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]
Figure 15
Figure 15. Figure 15: An overview of the 88 real-world objects in FillImpact, meticulously curated to capture the coupled acoustic dynamics of 3D geometry and varying [PITH_FULL_IMAGE:figures/full_fig_p012_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 3 linked inside Pith

  1. [1]

    That sounds right: Auditory self-supervision for dynamic robot manipulation,

    A. Thankaraj and L. Pinto, “That sounds right: Auditory self-supervision for dynamic robot manipulation,” inConference on Robot Learning. PMLR, 2023, pp. 1036–1049

    2023

  2. [2]

    Resound: interactive sound rendering for dynamic virtual environments,

    M. T. Taylor, A. Chandak, L. Antani, and D. Manocha, “Resound: interactive sound rendering for dynamic virtual environments,” inPro- ceedings of the 17th ACM international conference on Multimedia, 2009, pp. 271–280. 9

    2009

  3. [3]

    Visually indicated sounds,

    A. Owens, P. Isola, J. McDermott, A. Torralba, E. H. Adelson, and W. T. Freeman, “Visually indicated sounds,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 2405– 2413

    2016

  4. [4]

    Modal synthesis for vibrating objects,

    K. Van Den Doel, “Modal synthesis for vibrating objects,” 2003

    2003

  5. [5]

    Example-guided physically based modal sound synthesis,

    Z. Ren, H. Yeh, and M. C. Lin, “Example-guided physically based modal sound synthesis,”ACM Transactions on Graphics (TOG), vol. 32, no. 1, pp. 1–16, 2013

    2013

  6. [6]

    A dynamic analysis method of liquid-filled containers considering the fluid–structure interaction,

    X. Fang, X. Bao, F. Yue, and Q. Zhao, “A dynamic analysis method of liquid-filled containers considering the fluid–structure interaction,” Applied Sciences, vol. 14, no. 7, p. 2688, 2024

    2024

  7. [7]

    Dynamic analysis method of liquid-filled containers considering the fluid-structure interaction effect,

    X. bing Fang, X. Bao, F. Yue, and Q. Zhao, “Dynamic analysis method of liquid-filled containers considering the fluid-structure interaction effect,” 2024

    2024

  8. [8]

    Visual to sound: Generating natural sound for videos in the wild,

    Y . Zhou, Z. Wang, C. Fang, T. Bui, and T. L. Berg, “Visual to sound: Generating natural sound for videos in the wild,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 3550–3558

    2018

Show all 47 references
  1. [9]

    Action2sound: Ambient-aware generation of action sounds from egocentric videos,

    C. Chen, P. Peng, A. Baid, Z. Xue, W.-N. Hsu, D. Harwath, and K. Grau- man, “Action2sound: Ambient-aware generation of action sounds from egocentric videos,” inEuropean Conference on Computer Vision. Springer, 2024, pp. 277–295

    2024

  2. [10]

    Sonicsense: Object perception from in-hand acoustic vibration,

    J. Liu and B. Chen, “Sonicsense: Object perception from in-hand acoustic vibration,”arXiv preprint arXiv:2406.17932, 2024

    2024 arXiv

  3. [11]

    Physics- driven diffusion models for impact sound synthesis from videos,

    K. Su, K. Qian, E. Shlizerman, A. Torralba, and C. Gan, “Physics- driven diffusion models for impact sound synthesis from videos,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 9749–9759

    2023

  4. [12]

    Diffsound: Discrete diffusion model for text-to-sound generation,

    D. Yang, J. Yu, H. Wang, W. Wang, C. Weng, Y . Zou, and D. Yu, “Diffsound: Discrete diffusion model for text-to-sound generation,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 31, pp. 1720–1733, 2023

    2023

  5. [13]

    Learning neural acoustic fields,

    A. Luo, Y . Du, M. Tarr, J. Tenenbaum, A. Torralba, and C. Gan, “Learning neural acoustic fields,”Advances in Neural Information Processing Systems, vol. 35, pp. 3165–3177, 2022

    2022

  6. [14]

    Neraf: 3d scene infused neural radiance and acoustic fields,

    A. Brunetto, S. Hornauer, and F. Moutarde, “Neraf: 3d scene infused neural radiance and acoustic fields,”arXiv preprint arXiv:2405.18213, 2024

    2024

  7. [15]

    Soundspaces: Audio-visual navigation in 3d environments,

    C. Chen and U. Jain, “Soundspaces: Audio-visual navigation in 3d environments,” inECCV, 2020

    2020

  8. [16]

    Neural acoustic fields,

    X. Li, “Neural acoustic fields,”ACM Transactions on Graphics, 2022

    2022

  9. [17]

    3d gaussian splatting for real-time radiance field rendering,

    B. Kerbl, G. Kopanas, and T. Leimk ¨uhler, “3d gaussian splatting for real-time radiance field rendering,” inSIGGRAPH, 2023

    2023

  10. [18]

    Sonicgauss: 3d gaussian-based spatial sound generation,

    X. Wang, “Sonicgauss: 3d gaussian-based spatial sound generation,” arXiv preprint, 2025

    2025

  11. [19]

    Tangoflux: Diffusion-based text-to-audio generation,

    Y . Hung, “Tangoflux: Diffusion-based text-to-audio generation,”arXiv preprint, 2024

    2024

  12. [20]

    The theory of sound,

    J. W. S. Rayleigh, “The theory of sound,” 1896

  13. [21]

    Dover, 1945

    ——,The Theory of Sound. Dover, 1945

    1945

  14. [22]

    L. E. Kinsler,Fundamentals of Acoustics. Wiley, 2000

    2000

  15. [23]

    The use of helmholtz resonance for measuring the volume of liquids and solids,

    E. S. Webster and C. E. Davies, “The use of helmholtz resonance for measuring the volume of liquids and solids,”Sensors, vol. 10, no. 12, pp. 10 663–10 672, 2010

    2010

  16. [24]

    N. H. Fletcher and T. D. Rossing,The Physics of Musical Instruments. Springer, 1998

    1998

  17. [25]

    Granular solids, liquids, and gases,

    H. M. Jaeger and S. R. Nagel, “Granular solids, liquids, and gases,” Reviews of Modern Physics, 1996

    1996

  18. [26]

    Granular packings: nonlinear elasticity and sound prop- agation,

    H. A. Makse, “Granular packings: nonlinear elasticity and sound prop- agation,”Physical Review Letters, 2004

    2004

  19. [27]

    Sound-20k dataset,

    Y . Zhou, “Sound-20k dataset,” inCVPR, 2018

    2018

  20. [28]

    Objectfolder: A dataset of objects with implicit visual, auditory, and tactile representations,

    R. Gao, “Objectfolder: A dataset of objects with implicit visual, auditory, and tactile representations,” inCoRL, 2021

    2021

  21. [29]

    Objectfolder 2.0,

    ——, “Objectfolder 2.0,”IJRR, 2022

    2022

  22. [30]

    Realimpact: A dataset for spatial impact sound fields,

    X. Li, “Realimpact: A dataset for spatial impact sound fields,”IEEE Transactions on Multimedia, 2023

    2023

  23. [31]

    Hunyuan3d: A large-scale 3d generation model,

    L. et al., “Hunyuan3d: A large-scale 3d generation model,” 2024

    2024

  24. [32]

    glTF 2.0 Specification,

    Khronos 3D Formats Working Group, “glTF 2.0 Specification,” https: //github.com/KhronosGroup/glTF, 2021, accessed: 2026-03-14

    2021

  25. [33]

    Point transformer,

    H. Zhao, L. Jiang, J. Jia, P. H. Torr, and V . Koltun, “Point transformer,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 16 259–16 268

    2021

  26. [34]

    Splatformer: Point transformer for robust 3d gaussian splatting,

    Y . Chen, M. Mihajlovic, X. Chen, Y . Wang, S. Prokudin, and S. Tang, “Splatformer: Point transformer for robust 3d gaussian splatting,”arXiv preprint arXiv:2411.06390, 2024

    2024 arXiv

  27. [35]

    The physics of granular materials,

    H. M. Jaeger and S. R. Nagel, “The physics of granular materials,” Science, vol. 255, no. 5051, pp. 1523–1531, 1992

    1992

  28. [36]

    Stable audio open,

    Z. Evans, J. D. Parker, C. Carr, Z. Zukowski, J. Taylor, and J. Pons, “Stable audio open,” inICASSP 2025-2025 IEEE International Confer- ence on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2025, pp. 1–5

    2025

  29. [37]

    Fr ´echet audio distance: A reference-free metric for evaluating music enhancement algorithms,

    K. Kilgour, M. Zuluaga, D. Roblek, and M. Sharifi, “Fr ´echet audio distance: A reference-free metric for evaluating music enhancement algorithms,” inInterspeech, 2019

    2019

  30. [38]

    Cnn architectures for large-scale audio classification,

    S. Hershey, S. Chaudhuri, D. P. W. Ellis, J. F. Gemmeke, A. Jansen, R. C. Moore, M. Plakal, D. Platt, R. A. Saurous, B. Seybold, and M. Slaney, “Cnn architectures for large-scale audio classification,” inICASSP, 2017

    2017

  31. [39]

    On information and sufficiency,

    S. Kullback and R. A. Leibler, “On information and sufficiency,”Annals of Mathematical Statistics, vol. 22, no. 1, pp. 79–86, 1951

    1951

  32. [40]

    Improved techniques for training gans,

    T. Salimans, I. Goodfellow, W. Zaremba, V . Cheung, A. Radford, and X. Chen, “Improved techniques for training gans,” inNeurIPS, 2016

    2016

  33. [41]

    Recommendation p.800: Methods for subjective determination of transmission quality,

    ITU-T, “Recommendation p.800: Methods for subjective determination of transmission quality,” 1996

    1996

  34. [42]

    Crowdsourcing for speech quality evaluation,

    F. Ribeiro, D. Florencio, C. Zhang, and M. Seltzer, “Crowdsourcing for speech quality evaluation,” inICASSP, 2011

    2011

  35. [43]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” International Conference on Learning Representations (ICLR), 2019

    2019

  36. [44]

    Scaling instruction-finetuned language models,

    H. W. Chung, L. Hou, S. Longpre, B. Zoph, Y . Tay, W. Fedus, Y . Li, X. Wang, M. Dehghani, S. Brahmaet al., “Scaling instruction-finetuned language models,”arXiv preprint arXiv:2210.11416, 2022

    2022 arXiv

  37. [45]

    Exploring the limits of transfer learning with a unified text-to-text transformer,

    C. Raffel, N. Shazeer, A. Roberts, K. Lee, S. Narang, M. Matena, Y . Zhou, W. Li, and P. J. Liu, “Exploring the limits of transfer learning with a unified text-to-text transformer,”Journal of Machine Learning Research, vol. 21, no. 140, pp. 1–67, 2020. 10 APPENDIX A. Data Proc...

    2020

  38. [46]

    SonicGauss Fine-Tuning:The SonicGauss framework is natively designed to synthesize sound conditioned on 3D Gaussian Splatting representations and spatial impact coor- dinates. Consequently, we directly utilize the 3DGS models, precise strike positions, and the corresponding gr...

  39. [47]

    To leverage our entire dataset for fine-tuning TangoFlux, we systematically transform all multimodal physical parameters into rich textual prompts

    TangoFlux Fine-Tuning:Unlike models utilizing explicit 3D visual conditions, TangoFlux operates strictly as a text-to- audio generation framework. To leverage our entire dataset for fine-tuning TangoFlux, we systematically transform all multimodal physical parameters into rich...

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.