pith. sign in

arxiv: 2606.21372 · v1 · pith:PCN3DECOnew · submitted 2026-06-19 · 💻 cs.RO · cs.LG

NAC: Neural Action Codec for Vision-Language-Action Models

Ahad Jawaid , Yu Xiang This is my paper

Pith reviewed 2026-06-26 13:58 UTC · model grok-4.3

classification 💻 cs.RO cs.LG
keywords neural action codecaction tokenizationvision-language-action modelsRVQGANrobot manipulationautoregressive policieskinematic reconstruction
0
0 comments X

The pith

Neural Action Codec adapts audio-style RVQGANs to tokenize robot actions with lower error and higher success than binning or prior VQ methods.

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

The paper shows that short robot action trajectories can be compressed into discrete tokens by repurposing convolutional encoder-decoder architectures from neural audio codecs. It replaces audio-specific mel losses with simple time-domain and non-mel spectral reconstruction objectives, then uses offset codebooks and an ISTFT-based decoder with adversarial training to recover smooth trajectories. This produces a compact, ordered token space that standard autoregressive vision-language-action models can consume directly. Across simulation benchmarks and real manipulation tasks the resulting tokens yield lower reconstruction error and higher policy success rates than binning, FAST, and earlier vector-quantized approaches at comparable or better compression. The approach demonstrates that minimal changes to an established audio front-end suffice to create a practical action tokenizer for modern VLAs.

Core claim

NAC treats short robot action trajectories as multi-channel 1D signals and compresses them with a multi-scale RVQGAN whose encoder-decoder pair is drawn from neural audio codecs. Time-domain plus non-mel spectral losses replace mel-spectrogram objectives; offset codebooks produce an ordered discrete token sequence; a Vocos-style decoder with ISTFT head and adversarial discriminators reconstructs the original trajectory. The resulting tokenizer supplies VLAs with compact action sequences that support standard autoregressive decoding while preserving kinematic detail.

What carries the argument

Multi-scale RVQGAN with offset codebooks and ISTFT decoder, using time-domain and non-mel spectral losses to autoencode kinematic signals.

If this is right

  • Standard autoregressive VLAs can operate over the short structured token sequences produced by NAC.
  • The offset codebooks create a compact ordered token space without custom policy heads.
  • The ISTFT decoder recovers smooth, detailed trajectories suitable for low-level control.
  • NAC maintains or improves compression rates while raising downstream success on LIBERO-10, RoboMimic, and real tasks.

Where Pith is reading between the lines

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

  • The same loss substitution may allow audio-codec backbones to tokenize other continuous control signals such as joint torques or force profiles.
  • If the 1D-signal assumption holds for longer horizons, NAC-style tokenizers could reduce the sequence length needed for long-horizon planning.
  • Offset codebooks might be reused across different robot morphologies to share a common discrete action vocabulary.

Load-bearing premise

Short robot action trajectories behave like multi-channel 1D signals whose high-fidelity reconstruction requires only time-domain and non-mel spectral losses without major architectural redesign.

What would settle it

A controlled comparison on held-out real-world manipulation tasks in which NAC produces higher action reconstruction error or lower policy success rates than binning or FAST at matched token rates would falsify the performance claim.

Figures

Figures reproduced from arXiv: 2606.21372 by Ahad Jawaid, Yu Xiang.

Figure 1
Figure 1. Figure 1: Neural audio codecs [1, 2, 3] adapted for action tokenization. Top: Modern neural codecs compress raw waveforms into compact, multi-scale discrete codes, preserving coarse structures and fine temporal details. Bottom: NAC applies this approach to robot action chunks, treating actions as multi-channel 1D signals to learn a highly compressed, discrete latent space for downstream autoregressive policy learnin… view at source ↗
Figure 2
Figure 2. Figure 2: NAC overview: A continuous action chunk is flattened into a 1D pseudo-waveform and encoded by a SEANet-style encoder [37]. Multi-scale residual vector quantization [17] compresses the latent into discrete codes at progressively finer temporal resolutions. A Vocos-style decoder [38, 33] with an ISTFT head reconstructs the action chunk. The policy then models the resulting offset code sequence autoregressive… view at source ↗
Figure 3
Figure 3. Figure 3: Simulation environments. We benchmark tokenizers across both (a) a LIBERO-10 sub [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Real-world evaluation tasks. Outlines indicate initial start states (red) and successfully [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Full Real-world evaluation tasks. Outlines indicate the initial start states (red) and success [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
read the original abstract

Vision-language-action (VLA) models rely on discrete action tokenizers to bridge continuous robot control and autoregressive sequence modeling, yet existing tokenizers often trade off between compression, latency, and downstream performance. We revisit this design through the lens of neural audio codecs-convolutional encoder-decoder architectures with residual vector quantization that serve as the standard front end for audio foundation models. Motivated by their success, we introduce the Neural Action Codec (NAC), which treats short robot action trajectories as multi-channel 1D signals and compresses them using a multi-scale RVQGAN architecture. We observe that audio-specific mel-spectrogram objectives are ill-suited for kinematic signals; however, by replacing them with simple time-domain and non-mel spectral reconstruction losses, audio-codec-style models can autoencode actions with high fidelity without substantial architectural changes. NAC provides a compact, ordered token space via offset codebooks, enabling standard autoregressive policies to operate over short, structured sequences. Meanwhile, a Vocos-style decoder with an ISTFT head and adversarial discriminators recovers smooth, detailed trajectories. Across LIBERO-10, RoboMimic, and a suite of real-world manipulation tasks, NAC achieves lower reconstruction error and higher success rates than binning, FAST, and prior VQ-based tokenizers at comparable or better compression rates. These results demonstrate that repurposed neural audio codecs offer a strong, practical backbone for learned action tokenization in modern VLAs.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes the Neural Action Codec (NAC), which adapts multi-scale RVQGAN-style neural audio codecs to discretize short robot action trajectories by treating them as multi-channel 1D signals. It replaces audio-specific mel-spectrogram objectives with time-domain and non-mel spectral reconstruction losses, uses offset codebooks for ordered tokens, and employs a Vocos-style decoder with ISTFT head and adversarial discriminators. The central empirical claim is that NAC yields lower reconstruction error and higher success rates than binning, FAST, and prior VQ-based tokenizers on LIBERO-10, RoboMimic, and real-world manipulation tasks at comparable or better compression rates.

Significance. If the reported results hold, the work shows that minimal modifications to established audio codec architectures can produce effective action tokenizers for VLAs, enabling standard autoregressive policies over compact structured sequences. It explicitly credits the multi-channel 1D treatment, replacement of mel losses, and use of offset codebooks plus ISTFT/adversarial components for high-fidelity recovery without substantial redesign.

major comments (2)
  1. [Abstract and §4] The central performance claim (lower reconstruction error and higher success rates across LIBERO-10, RoboMimic, and real tasks) is load-bearing for the contribution, yet the provided abstract supplies no quantitative tables, error bars, statistical tests, or ablation details on loss components or codebook scales; full verification of the empirical wins requires these in the results section.
  2. [§3.1] §3.1: The multi-channel 1D signal treatment and choice of simple time-domain plus non-mel spectral losses are presented as sufficient without major redesign, but the manuscript should include an ablation isolating the contribution of each loss term to reconstruction fidelity to confirm they suffice for kinematic signals.
minor comments (2)
  1. [§3.2] Provide explicit equations or pseudocode for the non-mel spectral reconstruction loss to support reproducibility.
  2. [§4.1] Clarify how compression rates are matched exactly across all baselines (binning, FAST, prior VQ) in the experimental setup.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive review and recommendation of minor revision. The comments highlight opportunities to strengthen the empirical presentation, which we address point-by-point below with planned revisions.

read point-by-point responses

  1. Referee: [Abstract and §4] The central performance claim (lower reconstruction error and higher success rates across LIBERO-10, RoboMimic, and real tasks) is load-bearing for the contribution, yet the provided abstract supplies no quantitative tables, error bars, statistical tests, or ablation details on loss components or codebook scales; full verification of the empirical wins requires these in the results section.

    Authors: The results section (§4) already presents quantitative tables comparing reconstruction errors and task success rates for NAC against binning, FAST, and prior VQ tokenizers on LIBERO-10, RoboMimic, and real-world tasks at matched compression rates. To improve accessibility, we will revise the abstract to incorporate the key quantitative improvements (e.g., specific error reductions and success-rate gains) while respecting length constraints. We will also ensure error bars, confidence intervals, and any necessary statistical tests are included or added in the results tables. Ablation details on loss components and codebook scales will be expanded as part of the response to the second comment. revision: yes

  2. Referee: [§3.1] §3.1: The multi-channel 1D signal treatment and choice of simple time-domain plus non-mel spectral losses are presented as sufficient without major redesign, but the manuscript should include an ablation isolating the contribution of each loss term to reconstruction fidelity to confirm they suffice for kinematic signals.

    Authors: We agree that an explicit ablation isolating the individual contributions of the time-domain and non-mel spectral losses would provide stronger evidence that these terms suffice for kinematic signals without audio-specific mel objectives. In the revised manuscript we will add a dedicated ablation study (table or figure) quantifying reconstruction fidelity when each loss term is included or removed, confirming their combined effectiveness for action trajectories. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper frames NAC as an empirical architecture proposal that adapts existing RVQGAN-style audio codecs to multi-channel 1D action trajectories by swapping mel objectives for time-domain plus non-mel spectral losses. No equations, derivations, or predictions are presented that reduce to fitted parameters or self-citations by construction; the central claims rest on external benchmark comparisons (LIBERO-10, RoboMimic, real-world tasks) whose outcomes are not forced by the method definition itself. No load-bearing self-citation chains or uniqueness theorems appear in the abstract or described method.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central claim rests on treating actions as audio-like signals and on the sufficiency of the chosen reconstruction losses; these are domain assumptions rather than derived results. No free parameters are explicitly fitted in the abstract, but the model necessarily contains many neural-network weights and codebook sizes.

free parameters (1)
  • RVQ codebook sizes and scales
    Standard in RVQGAN architectures; values not stated in abstract but required for the tokenizer to function.
axioms (1)
  • domain assumption Action trajectories behave sufficiently like multi-channel 1D signals that audio-codec architectures transfer with only loss-function changes.
    Invoked in the abstract when motivating the replacement of mel-spectrogram objectives.
invented entities (1)
  • Neural Action Codec (NAC) no independent evidence
    purpose: Compact ordered tokenization of robot actions via adapted RVQGAN
    New model introduced in the paper; no independent evidence outside the reported experiments.

pith-pipeline@v0.9.1-grok · 5788 in / 1375 out tokens · 23561 ms · 2026-06-26T13:58:02.677832+00:00 · methodology

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

55 extracted references · 1 canonical work pages

  1. [1]

    Zeghidour, A

    N. Zeghidour, A. Luebs, A. Omran, J. Skoglund, and M. Tagliasacchi. SoundStream: An end-to- end neural audio codec.IEEE/ACM Transactions on Audio, Speech, and Language Processing, 30:495–507, 2021

    2021

  2. [2]

    Défossez, J

    A. Défossez, J. Copet, G. Synnaeve, and Y . Adi. High fidelity neural audio compression.Trans- actions on Machine Learning Research, 2023. ISSN 2835-8856. URL https://openreview. net/forum?id=ivCd8z8zR2. arXiv:2210.13438

    Pith/arXiv arXiv 2023

  3. [3]

    Kumar, P

    R. Kumar, P. Seetharaman, A. Luebs, I. Kumar, and K. Kumar. High-fidelity audio compression with improved RVQGAN. InAdvances in Neural Information Processing Systems, volume 36, pages 27980–27993, 2023. arXiv:2306.06546

    arXiv 2023

  4. [4]

    Black, N

    K. Black, N. Brown, D. Driess, A. Esmail, M. Equi, C. Finn, N. Fusai, L. Groom, K. Hausman, B. Ichter, et al. π0: A vision-language-action flow model for general robot control.arXiv preprint arXiv:2410.24164, 2024

    Pith/arXiv arXiv 2024

  5. [5]

    Black, N

    K. Black, N. Brown, J. Darpinian, K. Dhabalia, D. Driess, A. Esmail, M. R. Equi, C. Finn, N. Fusai, M. Y . Galliker, D. Ghosh, L. Groom, K. Hausman, B. Ichter, S. Jakubczak, T. Jones, L. Ke, D. LeBlanc, S. Levine, A. Li-Bell, M. Mothukuri, S. Nair, K. Pertsch, A. Z. Ren, L. X. Shi, L. Smith, J. T. Springenberg, K. Stachowicz, J. Tanner, Q. Vuong, H. Walke...

    2025

  6. [6]

    M. J. Kim, K. Pertsch, S. Karamcheti, T. Xiao, A. Balakrishna, S. Nair, R. Rafailov, E. P. Foster, P. R. Sanketi, Q. Vuong, T. Kollar, B. Burchfiel, R. Tedrake, D. Sadigh, S. Levine, P. Liang, and C. Finn. OpenVLA: An open-source vision-language-action model. In8th Annual Conference on Robot Learning, 2024. URLhttps://openreview.net/forum?id=ZMnD6QZAE6

    2024

  7. [7]

    C. Liu, X. Han, J. Gao, Y . Zhao, H. Chen, and Y . Du. OAT: Ordered action tokenization.arXiv preprint arXiv:2602.04215, 2026

    arXiv 2026

  8. [8]

    A. Yang, A. Li, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, C. Gao, C. Huang, C. Lv, et al. Qwen3 technical report.arXiv preprint arXiv:2505.09388, 2025

    Pith/arXiv arXiv 2025

  9. [9]

    Pertsch, K

    K. Pertsch, K. Stachowicz, B. Ichter, D. Driess, S. Nair, Q. Vuong, O. Mees, C. Finn, and S. Levine. FAST: Efficient action tokenization for vision-language-action models.arXiv preprint arXiv:2501.09747, 2025

    Pith/arXiv arXiv 2025

  10. [10]

    J. Lee, J. Duan, H. Fang, Y . Deng, S. Liu, B. Li, B. Fang, J. Zhang, Y . R. Wang, S. Lee, et al. MolmoAct: Action reasoning models that can reason in space.arXiv preprint arXiv:2508.07917, 2025

    Pith/arXiv arXiv 2025

  11. [11]

    D. Lee, C. Kim, S. Kim, M. Cho, and W.-S. Han. Autoregressive image generation using residual quantization. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11523–11532, 2022

    2022

  12. [12]

    Borsos, R

    Z. Borsos, R. Marinier, D. Vincent, E. Kharitonov, O. Pietquin, M. Sharifi, D. Roblek, O. Teboul, D. Grangier, M. Tagliasacchi, et al. AudioLM: A language modeling approach to audio generation.IEEE/ACM Transactions on Audio, Speech, and Language Processing, 31:2523– 2533, 2023

    2023

  13. [13]

    C. Wang, S. Chen, Y . Wu, Z. Zhang, L. Zhou, S. Liu, Z. Chen, Y . Liu, H. Wang, J. Li, et al. Neural codec language models are zero-shot text to speech synthesizers.arXiv preprint arXiv:2301.02111, 2023. 9

    Pith/arXiv arXiv 2023

  14. [14]

    Cadene, S

    R. Cadene, S. Alibert, F. Capuano, M. Aractingi, A. Zouitine, P. Kooijmans, J. Choghari, M. Russi, C. Pascal, S. Palma, D. Aubakirova, M. Shukor, J. Moss, A. Soare, Q. Lhoest, Q. Gallouédec, and T. Wolf. Lerobot: An open-source library for end-to-end robot learning. InThe Fourteenth International Conference on Learning Representations, 2026. URLhttps: //o...

    2026

  15. [15]

    Kaneko, K

    T. Kaneko, K. Tanaka, H. Kameoka, and S. Seki. iSTFTNet: Fast and lightweight mel- spectrogram vocoder incorporating inverse short-time fourier transform. InICASSP 2022–2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 6207–6211. IEEE, 2022

    2022

  16. [16]

    S. S. Stevens, J. V olkmann, and E. B. Newman. A scale for the measurement of the psychological magnitude pitch.The journal of the acoustical society of america, 8(3):185–190, 1937

    1937

  17. [17]

    Siuzdak, F

    H. Siuzdak, F. Grötschla, and L. A. Lanzendörfer. SNAC: Multi-scale neural audio codec. InAudio Imagination: NeurIPS 2024 Workshop on AI-Driven Speech, Music, and Sound Generation, 2024. URLhttps://openreview.net/forum?id=PFBF5ctj4X

    2024

  18. [18]

    D. Yang, J. Tian, X. Tan, R. Huang, S. Liu, X. Chang, J. Shi, S. Zhao, J. Bian, Z. Zhao, et al. UniAudio: An audio foundation model toward universal audio generation.arXiv preprint arXiv:2310.00704, 2023

    arXiv 2023

  19. [19]

    Beyer, A

    L. Beyer, A. Steiner, A. S. Pinto, A. Kolesnikov, X. Wang, D. Salz, M. Neumann, I. Alab- dulmohsin, M. Tschannen, E. Bugliarello, T. Unterthiner, D. Keysers, S. Koppula, F. Liu, A. Grycner, A. Gritsenko, N. Houlsby, M. Kumar, K. Rong, J. Eisenschlos, R. Kabra, M. Bauer, M. Bošnjak, X. Chen, M. Minderer, P. V oigtlaender, I. Bica, I. Balazevic, J. Puigcerv...

    Pith/arXiv arXiv 2024

  20. [20]

    Touvron, L

    H. Touvron, L. Martin, K. Stone, P. Albert, A. Almahairi, Y . Babaei, N. Bashlykov, S. Batra, P. Bhargava, S. Bhosale, D. Bikel, L. Blecher, C. C. Ferrer, M. Chen, G. Cucurull, D. Esiobu, J. Fernandes, J. Fu, W. Fu, B. Fuller, C. Gao, V . Goswami, N. Goyal, A. Hartshorn, S. Hosseini, R. Hou, H. Inan, M. Kardas, V . Kerkez, M. Khabsa, I. Kloumann, A. Koren...

    Pith/arXiv arXiv 2023

  21. [21]

    H. Fang, J. Duan, D. Clay, S. Wang, S. Liu, W. Huang, X. Fan, W.-C. Tsai, S. Chen, Y . R. Wang, S. Xing, J. Cho, J. S. Park, A. Eftekhar, P. Sushko, K. Farley, A. Wadhwa, C. Harrison, W. Han, Y .-C. Lee, E. VanderBilt, R. Hendrix, S. Ellawela, L. Ngoo, J. Chai, Z. Ren, A. Farhadi, D. Fox, and R. Krishna. Molmoact2: Action reasoning models for real-world d...

    Pith/arXiv arXiv 2026

  22. [22]

    Intelligence, B

    P. Intelligence, B. Ai, A. Amin, R. Aniceto, A. Balakrishna, G. Balke, K. Black, G. Bokin- sky, S. Cao, T. Charbonnier, V . Choudhary, F. Collins, K. Conley, G. Connors, J. Darpinian, K. Dhabalia, M. Dhaka, J. DiCarlo, D. Driess, M. Equi, A. Esmail, Y . Fang, C. Finn, C. Glos- sop, T. Godden, I. Goryachev, L. Groom, H. Habeeb, H. Hancock, K. Hausman, G. H...

    Pith/arXiv arXiv 2026

  23. [23]

    Ahmed, T

    N. Ahmed, T. Natarajan, and K. R. Rao. Discrete cosine transform.IEEE transactions on Computers, 100(1):90–93, 1974

    1974

  24. [24]

    Sennrich, B

    R. Sennrich, B. Haddow, and A. Birch. Neural machine translation of rare words with subword units. InProceedings of the 54th annual meeting of the association for computational linguistics (volume 1: long papers), pages 1715–1725, 2016

    2016

  25. [25]

    Y . Wang, H. Zhu, M. Liu, J. Yang, H.-S. Fang, and T. He. VQ-VLA: Improving vision-language- action models via scaling vector-quantized action tokenizers. InProceedings of the IEEE/CVF International Conference on Computer Vision, 2025. arXiv:2507.01016

    arXiv 2025

  26. [26]

    R. Gray. Vector quantization.IEEE Assp Magazine, 1(2):4–29, 1984

    1984

  27. [27]

    Z. Dong, Y . Liu, S. Zhang, B. Ye, Y . Yuan, F. Ni, J. Gong, X. Qiu, H. Zhao, Y . Li, et al. ActionCodec: What makes for good action tokenizers.arXiv preprint arXiv:2602.15397, 2026

    arXiv 2026

  28. [28]

    Y . Liu, S. Zhang, Z. Dong, B. Ye, T. Yuan, X. Yu, L. Yin, C. Lu, J. Shi, L. J.-T. Yu, L. Zheng, J. Gong, T. Jiang, X. Qiu, and H. Zhao. FASTer: Toward powerful and efficient autoregressive vision–language–action models with learnable action tokenizer and block-wise decoding. In The Fourteenth International Conference on Learning Representations, 2026. UR...

    arXiv 2026

  29. [29]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017

    2017

  30. [30]

    H. Zhao, O. Gallo, I. Frosio, and J. Kautz. Loss functions for image restoration with neural networks.IEEE Transactions on computational imaging, 3(1):47–57, 2016

    2016

  31. [31]

    Rippel, M

    O. Rippel, M. Gelbart, and R. Adams. Learning ordered representations with nested dropout. In International Conference on Machine Learning, pages 1746–1754. PMLR, 2014

    2014

  32. [32]

    Darcet, M

    T. Darcet, M. Oquab, J. Mairal, and P. Bojanowski. Vision transformers need registers. In International Conference on Learning Representations (ICLR), 2024

    2024

  33. [33]

    H. Siuzdak. V ocos: Closing the gap between time-domain and fourier-based neural vocoders for high-quality audio synthesis.arXiv preprint arXiv:2306.00814, 2023

    arXiv 2023

  34. [34]

    L. Chen, K. Lu, A. Rajeswaran, K. Lee, A. Grover, M. Laskin, P. Abbeel, A. Srinivas, and I. Mordatch. Decision transformer: reinforcement learning via sequence modeling. InProceed- ings of the 35th International Conference on Neural Information Processing Systems, pages 15084–15097, 2021

    2021

  35. [35]

    T. Z. Zhao, V . Kumar, S. Levine, and C. Finn. Learning fine-grained bimanual manipulation with low-cost hardware.arXiv preprint arXiv:2304.13705, 2023

    Pith/arXiv arXiv 2023

  36. [36]

    Torabi, G

    F. Torabi, G. Warnell, and P. Stone. Behavioral cloning from observation. InProceedings of the 27th International Joint Conference on Artificial Intelligence, pages 4950–4957, 2018

    2018

  37. [37]

    Tagliasacchi, Y

    M. Tagliasacchi, Y . Li, K. Misiunas, and D. Roblek. SEANet: A Multi-Modal Speech En- hancement Network. InInterspeech 2020, pages 1126–1130, 2020. doi:10.21437/Interspeech. 2020-1563

    work page doi:10.21437/interspeech 2020

  38. [38]

    J. Kong, J. Kim, and J. Bae. HiFi-GAN: Generative adversarial networks for efficient and high fidelity speech synthesis. InAdvances in Neural Information Processing Systems, volume 33, pages 17022–17033, 2020. 11

    2020

  39. [39]

    Kiranyaz, O

    S. Kiranyaz, O. Avci, O. Abdeljaber, T. Ince, M. Gabbouj, and D. J. Inman. 1d convolutional neural networks and applications: A survey.Mechanical systems and signal processing, 151: 107398, 2021

    2021

  40. [40]

    K. He, X. Zhang, S. Ren, and J. Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770– 778, 2016

    2016

  41. [41]

    Clevert, T

    D.-A. Clevert, T. Unterthiner, and S. Hochreiter. Fast and accurate deep network learning by exponential linear units (elus).arXiv preprint arXiv:1511.07289, 4(5):11, 2015

    Pith/arXiv arXiv 2015

  42. [42]

    Salimans and D

    T. Salimans and D. P. Kingma. Weight normalization: A simple reparameterization to accelerate training of deep neural networks.Advances in neural information processing systems, 29, 2016

    2016

  43. [43]

    Isola, J.-Y

    P. Isola, J.-Y . Zhu, T. Zhou, and A. A. Efros. Image-to-image translation with conditional adversarial networks. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 1125–1134, 2017

    2017

  44. [44]

    MacQueen

    J. MacQueen. Multivariate observations. InProceedings ofthe 5th Berkeley symposium on mathematical statisticsand probability, volume 1, pages 281–297. University of California press Oakland, CA, USA, 1967

    1967

  45. [45]

    Van Den Oord, O

    A. Van Den Oord, O. Vinyals, et al. Neural discrete representation learning.Advances in neural information processing systems, 30, 2017

    2017

  46. [46]

    A. Roy, A. Vaswani, A. Neelakantan, and N. Parmar. Theory and experiments on vector quantized autoencoders.arXiv preprint arXiv:1805.11063, 2018

    Pith/arXiv arXiv 2018

  47. [47]

    S. Targ, D. Almeida, and K. Lyman. Resnet in resnet: Generalizing residual architectures.arXiv preprint arXiv:1603.08029, 2016

    Pith/arXiv arXiv 2016

  48. [48]

    Z. Liu, H. Mao, C.-Y . Wu, C. Feichtenhofer, T. Darrell, and S. Xie. A convnet for the 2020s. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11976–11986, 2022

    2022

  49. [49]

    I. J. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio. Generative adversarial nets.Advances in neural information processing systems, 27, 2014

    2014

  50. [50]

    W. Jang, D. Lim, J. Yoon, B. Kim, and J. Kim. UnivNet: A neural vocoder with multi- resolution spectrogram discriminators for high-fidelity waveform generation.arXiv preprint arXiv:2106.07889, 2021

    arXiv 2021

  51. [51]

    J. Shen, R. Pang, R. J. Weiss, M. Schuster, N. Jaitly, Z. Yang, Z. Chen, Y . Zhang, Y . Wang, R. Skerrv-Ryan, et al. Natural tts synthesis by conditioning wavenet on mel spectrogram predictions. In2018 IEEE international conference on acoustics, speech and signal processing (ICASSP), pages 4779–4783. IEEE, 2018

    2018

  52. [52]

    Radford, K

    A. Radford, K. Narasimhan, T. Salimans, I. Sutskever, et al. Improving language understanding by generative pre-training.Openai, 2018

    2018

  53. [53]

    B. Liu, Y . Zhu, C. Gao, Y . Feng, Q. Liu, Y . Zhu, and P. Stone. Libero: Benchmarking knowledge transfer for lifelong robot learning.Advances in Neural Information Processing Systems, 36: 44776–44791, 2023

    2023

  54. [54]

    Mandlekar, D

    A. Mandlekar, D. Xu, J. Wong, S. Nasiriany, C. Wang, R. Kulkarni, L. Fei-Fei, S. Savarese, Y . Zhu, and R. Martín-Martín. What matters in learning from offline human demonstrations for robot manipulation. InConference on Robot Learning, pages 1678–1690. PMLR, 2022

    2022

  55. [55]

    C. Chi, Z. Xu, S. Feng, E. Cousineau, Y . Du, B. Burchfiel, R. Tedrake, and S. Song. Diffusion policy: Visuomotor policy learning via action diffusion.The International Journal of Robotics Research, 44(10-11):1684–1704, 2025. arXiv:2303.04137. 12 Appendix A Method Details A.1 Autoregressive Policy Inference Algorithm 2Autoregressive NAC Policy Inference R...

    Pith/arXiv arXiv 2025