Pith.

Robotics · arXiv:2608.27384 v1 · submitted 27 Aug 2026 · 32 references

FlashVLA: Streaming Action Decoding for Fast and Asynchronous VLA Inference

Zekai Li, Jiaming Tang, Zhijian Liu

The paper says

“Vision-Language-Action (VLA) models are increasingly promising for robotic manipulation, yet their real-world deployment remains bottlenecked by high inference latency and unstable asynchronous execution.”

more of the abstract

This challenge is particularly pronounced in flow-matching-based VLA models, where action decoding requires multiple iterative steps conditioned on the VLM context. While efficient inference methods improve control frequency and asynchronous methods reduce execution idle time, existing approaches often fail to jointly achieve low-latency inference and accurate, temporally consistent asynchronous execution. We introduce \textbfFlashVLA, a streaming action decoding framework that addresses both challenges in a unified formulation. FlashVLA maintains a streaming action buffer with multiple chunks at different noise levels and decodes them using chunk-wise causal attention. This design allows FlashVLA to produce one executable action chunk per inference step. Moreover, its chunk-wise autoregressive formulation implicitly preserves action continuity, enabling smooth asynchronous execution without extra future-state conditioning. Across extensive simulated and real-world experiments, FlashVLA substantially improves inference speed while maintaining strong task performance. It can achieve ≥30\,Hz control frequency on a single GPU with smooth asynchronous inference in real-world deployment.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 9 Sep 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

The passes disagreed. Reader: holds up. Skeptic: agrees with the reader. Referee: major revision.

Clean engineering paper that streams diffusion across action chunks for fast, asynchronous VLA inference; the core idea is adapted from video streaming diffusion, but the adaptation to robot action decoding is careful and empirically well-supported. This line is wrong

2 major objections · 6 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 32 references say.

The paper says, at §3.1, footnote 1, Table 2

The review did not quote this passage.

Pith says: Objection 1 of 2 · major

Please reconcile the headline efficiency claims with the matched latency measurements. The abstract, §1, and §3.1 state that 'per-step action-decoding latency drops by up to 20×', and footnote 1 attributes this to a 'ten-to-one pass reduction' yielding a '9.3× reduction in GPU work' and a measured '19.9×' wall-clock speedup.

more of objection 1

With the reported default configurations (buffer length N=4 on LIBERO and RoboTwin 2.0, N=5 on SmolVLA), each chunk traverses N noise levels, so the pass reduction relative to π0.5's ten denoising steps is at most 2.5×, not 10×. The 9.3× GPU-work claim would additionally require a four-chunk causal buffer pass to cost roughly 1.08× a single-chunk pass, which the data in Table 2 appear to contradict: if π0.5's 45.8 ms total contains the stated 75% action decoding (≈34.4 ms over ten passes, ≈3.4 ms/pass), then FlashVLA's 26.7 ms total with the same VLM portion implies a buffer pass of ≈15.3 ms, about 4.5× a single π0.5 pass, and a per-chunk decoder speedup of only ≈2.2×. In addition, the 3.5× overall speedup cited for Fig. 1b is not consistent with the 1.72× latency reduction in Table 2. Please report the profiled components of Fig. 1b and Table 2 on the same basis (same GPU, views, system optimizations, buffer configuration), and state explicitly which microbenchmark conditions produce the 19.9×/20× numbers, or reword the abstract and conclusion to the matched end-to-end figures.

The paper says, at §4.4, Table 4

The review did not quote this passage.

Pith says: Objection 2 of 2 · major

The 36.6-point long-horizon gain on RoboTwin 2.0 (Clean 54.2→90.8, Random 51.8→88.4) is attributed in §4.4 to 'chunk-level memory' and stated to 'follow directly from the chunk-wise causal formulation.' As presented, this attribution is not isolated from other differences between FlashVLA and π0.5 in that comparison: FlashVLA's action expert receives enhanced FiLM multi-level timestep conditioning (Appendix A.4), its training objective (multi-buffer …

more of objection 2

… packing, Eq. 3) differs from π0.5's independent-chunk objective, norm layers are reinitialized under the gradient-explosion fix for the 50-task multitask run, and the buffer span (20×4=80) is not matched to π0.5's native chunk length of 50. The causal-mask ablation in Fig. 6 is run on LIBERO and reports asynchronous success rate; it does not provide a control for the synchronous long-horizon gain on RoboTwin 2.0. Please provide (a) the no-causal control with the same buffer, same multi-buffer fine-tuning, and same FiLM conditioning evaluated on the RoboTwin horizon groups, (b) multiple seeds or reported variance for the 50-task training runs (the appendix states one model per method), and (c) an error decomposition showing where π0.5 fails on long-horizon tasks. Without such controls, the causal claim in §4.4 should be softened to a hypothesis.

6 minor objections

The paper says, at §3.1, Eq. (2)

Pith says: minor objection 1 of 6

The definition asserts τ_1 < τ_2 < ⋯ < τ_N, but independent Beta perturbations u_i do not guarantee this ordering for every draw; please state that the staircase is sorted after sampling or that the ordering holds in expectation.

The paper says, at §3.3, Eq. (3)

Pith says: minor objection 2 of 6

Equation (3) sums the flow-matching loss over all i ≤ j without specifying how padding chunks are handled; please state explicitly that the loss is masked over padded slots.

show 4 more

The paper says, at §4.2, first paragraph

Pith says: minor objection 3 of 6

The sentence 'We used to denote the number of low-level action steps...' appears to be missing the symbol d; please introduce d explicitly before it is used in Table 1.

The paper says, at §4.4, Tables 3–4

Pith says: minor objection 4 of 6

LIBERO results are reported as averages over five runs (Appendix A.1), but no seed count or variance is stated for the RoboTwin 2.0 results, which matters in light of the 36.6-point gap; please state the number of seeds or single-run status in the main text.

The paper says, at §5, Fig. 5

Pith says: minor objection 5 of 6

The real-robot evaluation uses 15 trials per task with a three-point rubric, and the 'smooth 30 Hz' claim rests on a mean inference latency of 67.3 ms, which is marginally above two nominal 30 Hz control periods (66.7 ms); please report per-task score variance and the measured timing distribution or jitter used to verify smooth 30 Hz operation.

The paper says, at Table 1

Pith says: minor objection 6 of 6

The StreamingVLA rows report 'n/a' for Steps and Time; please state why these were not measured or provide the values for completeness.

What would settle it

To further validate the robustness of the fine-tuning adaptation, run the same multi-buffer joint fine-tuning protocol on a fourth flow-matching VLA architecture (e.g., GR00T N1) using the LIBERO benchmark. If the success rate remains within 2 points of the baseline π0.5 result, the adaptation assumption is robust. If it drops significantly, the method may require architecture-specific tuning beyond what is documented.

from the skeptic's reading

How the review went

4 readings by deepseek-v4-flash · 9 Sep 2026

  1. Reader holds up confidence high
  2. Skeptic agrees 1 min 1 s · 16,470 tokens read, 6,568 written
  3. Editor wrote the verdict above 11 s · 17,168 tokens read, 980 written
  4. Referee major revision: 2 major objections, 6 minor 3 min 13 s · 16,864 tokens read, 21,691 written

Not run on this paper: the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The review in brief, machine-written

This paper does one thing and does it thoroughly. FlashVLA takes the streaming chunk-wise diffusion idea that has been working well in long-video generation (Diffusion Forcing, StreamDit, Magi-1) and adapts it to robot action decoding in flow-matching VLAs. The adaptation is not trivial: they match buffer depth and noise schedule to control-frequency targets, apply causal masking at chunk granularity, and design a multi-buffer joint fine-tuning that packs all cold-start configurations into a single training sample to avoid dataset blowup. The result is a drop-in modification that simultaneously reduces latency (by amortizing denoising steps across time) and recovers asynchronous continuity without an explicit future-state predictor.

The experimental support is the paper's strongest asset. The authors evaluate on two major benchmarks (LIBERO, RoboTwin 2.0) and a real Franka arm, across three different base VLAs (π0.5, SmolVLA, LingBot-VLA). They show consistent latency reductions (up to 2.4× end-to-end speedup) and success rates that match or exceed the synchronous baseline under asynchronous delays up to d=4. The ablation on the causal mask (Fig. 6) cleanly isolates the active ingredient. The cross-architecture results show the method is not tied to a single codebase.

Soft spots: the 36.6-point long-horizon gain on RoboTwin 2.0 (Table 4) is startlingly large and could benefit from an analysis of whether the baseline π0.5 has a particular failure mode on those tasks (e.g., losing track of where it is in a long sequence) that FlashVLA's chunk-level memory trivially fixes. The paper discusses this as a structural advantage, and the causal-mask ablation provides indirect support, but a direct diagnostic (e.g., measuring chunk-level state estimation) would raise confidence. Also, the method relies on fine-tuning pretrained models; while the authors demonstrate it works on three architectures, a pretrained-from-scratch VLA under the chunk-wise causal objective is flagged as future work and might yield further gains. These are minor concerns relative to the paper's overall empirical strength.

The paper is well-written, the algorithmic description is reproducible, and the contribution is practically significant for deploying VLAs on real robots. It does not claim paradigm-shifting novelty, but it solves a real bottleneck with a clean design. I would bring it to a reading group if people work on robot learning or efficient generative models. I would cite it for the streaming buffer + chunk-wise causal attention recipe.

Recommendation: publish after the authors add some diagnostic analysis on the long-horizon gain to confirm it is not a baseline artifact. No review-blocking issues.

The strongest objection

The paper presents a well-supported method. The central claim that a streaming buffer with chunk-wise causal attention simultaneously reduces latency and preserves asynchronous continuity is empirically validated across multiple benchmarks, architectures, and a real robot. The reader's identified weakest assumption—that the multi-buffer joint fine-tuning adapts pretrained models without retraining from scratch—is addressed by experiments on three diverse flow-matching VLAs, with a documented fix for gradient instability (norm reinitialization). No internal inconsistencies or unsupported leaps were found. The empirical evidence is sufficient to support the claim, and no additional load-bearing concern emerges from a careful read.

The reader accurately identified the weakest assumption: the multi-buffer joint fine-tuning must adapt pretrained models to the streaming paradigm. However, the paper provides strong empirical support on three architectures (π0.5, SmolVLA, LingBot-VLA) and a real robot, with ablation studies isolating the active ingredient and robustness across hyperparameters. The gradient explosion fix is documented and does not undermine the central claim. The concern is not strong enough to change the verdict because the evidence directly addresses it: the method works on the tested models, and the claim does not require universal architecture coverage. Therefore, the original ACCEPT verdict remains appropriate.

The referee's report

If the results hold, FlashVLA is a practically significant advance: it addresses decoding latency and asynchronous prediction-execution mismatch with one mechanism that can be attached to existing flow-matching VLA checkpoints by fine-tuning, without step-count distillation or a new future-state module. The empirical base is broad: two simulators with horizon-grouped analysis, three model families, latency and reaction-time benchmarks against dedicated system-level baselines, a real-robot deployment, and a causal-mask ablation (Fig. 6) that directly tests the proposed mechanism. The paper ships a code link and documents a concrete training-stability fix (norm-layer reinitialization, Appendix A.4), which strengthens reproducibility. The cross-architecture results (Table 5) are credible evidence against the main risk of the approach, namely that pretrained action experts might not adapt to chunk-wise causal attention without retraining from scratch.

Pith's reading of the paper (a machine's summary, not the authors')

This paper argues that the two dominant bottlenecks in deploying vision-language-action (VLA) models for real-time robot control—high inference latency and prediction-execution mismatch under asynchronous execution—share a single structural cause: decoding each action chunk in isolation. FlashVLA replaces isolated chunk decoding with a streaming buffer that holds several action chunks at different noise levels and advances them jointly using chunk-wise causal attention. Each forward pass progresses every chunk by one denoising step, amortizing the denoising cost across time and emitting one executable chunk per step. Because noisier future chunks attend to cleaner near-execution chunks, the model implicitly conditions on the ongoing trajectory, recovering smooth asynchronous continuity without a separate future-state predictor. The result is up to 20× lower per-step action-decoding latency and sustained ≥30 Hz control on a single GPU, with task success matching or exceeding the synchronous baseline across simulated and real-world benchmarks.

What carries the argument. The key machinery is a streaming action buffer of N chunks at staggered noise levels (τ₁ < τ₂ < … < τ_N) combined with a chunk-wise causal attention mask. The buffer is maintained as a queue: after warm-up, each forward pass advances all chunks by one denoising step, pops the cleanest chunk for execution, shifts the remaining chunks forward, and appends a fresh pure-noise chunk at the tail. The chunk-wise causal mask ensures that noisier (future) chunks attend only to cleaner (near-execution) chunks, establishing temporal information flow forward from the trajectory being executed. This design amortizes the N denoising steps across N forward passes (each step also advances every other chunk), and implicitly conditions future action predictions on the robot's ongoing motion, removing the need for explicit mismatch correction.

Core claim. The central discovery is that a single design choice—a streaming buffer of action chunks at staggered noise levels with chunk-wise causal attention—simultaneously eliminates the latency cost and the asynchronous-mismatch cost that prior work treated as separate problems. In flow-matching VLAs, each chunk decode previously required ten sequential denoising steps concentrated in one inference pass (the latency cost), and each new chunk was decoded without knowledge of chunks already in flight (the mismatch cost). FlashVLA's streaming buffer spreads the ten denoising steps across separate forward passes, so that one denoised chunk is produced per step. The chunk-wise causal mask lets future (noisier) chunks attend to earlier (cleaner) chunks, giving the decoder a structured view of the trajectory being executed. This implicit conditioning replaces explicit future-state prediction or action-conditioning modules. The paper demonstrates that this single mechanism achieves 2.43× end-to-end per-action speedup on LIBERO while improving average success rate from 96.9% to 97.8%, and generalizes to RoboTwin 2.0 and real-world Franka tasks at 30 Hz control.

The premise it rests on. The pretrained flow-matching VLA's action expert can be adapted to handle chunk-wise causal attention with multi-level timestep conditioning via multi-buffer joint fine-tuning, without requiring retraining from scratch or suffering generalization failure on unseen architectures.

If the paper is right (Pith's inferences, not the authors')

Flow-matching VLAs can achieve real-time control at ≥30 Hz on a single consumer GPU without sacrificing task success, removing a key deployment barrier for robotics.

The streaming buffer design is a drop-in modification for any pretrained flow-matching VLA, requiring only a multi-buffer joint fine-tuning pass to adapt the action expert to chunk-wise causal attention.

Asynchronous execution of VLA models becomes lossless under chunk-wise causal attention: FlashVLA maintains success above the synchronous baseline even as the prediction-execution delay (d) grows to 4 steps, whereas prior asynchronous methods degrade sharply.

Long-horizon tasks benefit disproportionately from the streaming buffer's implicit chunk-level memory: FlashVLA improves success on RoboTwin 2.0 long-horizon tasks by 36.6 percentage points over the baseline under synchronous evaluation.

The streaming buffer idea could be extended to other iterative generative policies beyond flow matching (e.g., diffusion policies), provided the chunk-wise causal attention pattern is preserved. This would broaden the scope of the method to a larger class of action decoders.

Because FlashVLA's streaming inference structure is computationally uniform across steady-state steps, it is a natural target for hardware-specific compilation (e.g., FPGA or TPU pipelines), potentially further reducing latency below the GPU figures reported.

The multi-buffer joint fine-tuning strategy packs all cold-start buffer configurations into one sample; this packing insight might be applicable to other streaming architectures that need to handle variable-length buffer states during training.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper · Review history (2 versions)

The paper

Abstract

Vision-Language-Action (VLA) models are increasingly promising for robotic manipulation, yet their real-world deployment remains bottlenecked by high inference latency and unstable asynchronous execution. This challenge is particularly pronounced in flow-matching-based VLA models, where action decoding requires multiple iterative steps conditioned on the VLM context. While efficient inference methods improve control frequency and asynchronous methods reduce execution idle time, existing approaches often fail to jointly achieve low-latency inference and accurate, temporally consistent asynchronous execution. We introduce \textbfFlashVLA, a streaming action decoding framework that addresses both challenges in a unified formulation. FlashVLA maintains a streaming action buffer with multiple chunks at different noise levels and decodes them using chunk-wise causal attention. This design allows FlashVLA to produce one executable action chunk per inference step. Moreover, its chunk-wise autoregressive formulation implicitly preserves action continuity, enabling smooth asynchronous execution without extra future-state conditioning. Across extensive simulated and real-world experiments, FlashVLA substantially improves inference speed while maintaining strong task performance. It can achieve ≥30\,Hz control frequency on a single GPU with smooth asynchronous inference in real-world deployment.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-09-23

The authors' main claim, as the classifier picked it out 0.99

“Across extensive simulated and real-world experiments, FlashVLA substantially improves inference speed while maintaining strong task performance.”

How the main claim is stated: asserted 1.00

What the abstract itself states, as probabilities: claims first 0.07 · states limitation 0.10 · human participants 0.06 · controlled comparison 0.17 · randomized assignment 0.01 · states quantitative result 0.98 · compares with prior results 0.55 · mentions code or data availability 0.03

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
Vision-Language-Action (VLA) models are increasingly promising for robotic manipulation, yet their real-world deployment remains bottlenecked by high inference latency and unstable asynchronous execution. background 0.97 0.06 0.03 asserted 0.94
This challenge is particularly pronounced in flow-matching-based VLA models, where action decoding requires multiple iterative steps conditioned on the VLM context. background 0.99 0.07 0.02 asserted 0.99
While efficient inference methods improve control frequency and asynchronous methods reduce execution idle time, existing approaches often fail to jointly achieve low-latency inference and accurate, temporally consistent asynchronous execution. gap 0.95 0.06 0.05 asserted 0.96
We introduce \textbf{FlashVLA}, a streaming action decoding framework that addresses both challenges in a unified formulation. contribution 1.00 0.86 0.15 asserted 1.00
FlashVLA maintains a streaming action buffer with multiple chunks at different noise levels and decodes them using chunk-wise causal attention. method 0.98 0.80 0.05 asserted 1.00
This design allows FlashVLA to produce one executable action chunk per inference step. method 0.93 0.81 0.07 asserted 0.97
Moreover, its chunk-wise autoregressive formulation implicitly preserves action continuity, enabling smooth asynchronous execution without extra future-state conditioning. method 0.78 0.75 0.09 asserted 0.96
Across extensive simulated and real-world experiments, FlashVLA substantially improves inference speed while maintaining strong task performance. result 1.00 0.94 0.62 asserted 1.00
It can achieve $\geq$30\,Hz control frequency on a single GPU with smooth asynchronous inference in real-world deployment. result 1.00 0.93 0.26 asserted 0.77

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.98

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 6,834 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (8)
Figure 1 of the paper
Figure 1. (a) Two prevailing VLA inference patterns and FlashVLA’s joint solution. (b) Per-policy
Figure 2 of the paper
Figure 2. Overview of FlashVLA. Left: Multi-buffer joint fine-tuning packs all cold-start configu￾rations for one observation into a single sample. Attention masks isolate buffers, and chunk-wise causal attention lets noisier chunks attend only to cleaner ones. Right: Streaming inference runs the buffer as a queue: cold-start fills it with N−1 warm-up passes, while steady streaming advances all chunks, executes the cleanest one, and appends fresh noise. Gray squares denote padding. with denoising progress and with execution order simultaneously, so the chunk most advanced in denoising is also…
Figure 3 of the paper
Figure 3. Success rate and time per step across asynchronous delay
Figure 4 of the paper
Figure 4. Reaction speed under a 30 Hz target. FlashVLA achieves 1.7× lower time-to-first￾action (TTFA) and 1.6× lower time-to-react (TTR) than FASTER. Method LIBERO RoboTwin 2.0 Spatial Object Goal Long Avg. Clean Random Avg. π0.5 98.8 98.2 98.0 92.4 96.9 86.1 85.8 86.0 +FlashVLA 98.6 99.0 97.8 96.2 97.9 (↑ 1.0) 90.8 90.2 90.5 (↑ 4.5)
Figure 5 of the paper
Figure 5. Real-world tasks and results. Top: the short-, medium-, and long-horizon tasks used
Figure 6 of the paper
Figure 6. Chunk-wise causal attention ablation on LIBERO. Removing the causal mask while keeping
Figure 7 of the paper
Figure 7. Chunk size C ablation on LIBERO with buffer length N=5 fixed. (a) Success rate. (b) Per-step latency. Task Pick and Place Whiteboard Wiping Table Cleaning Instruction Pick up the fruit and place it into the basket Wipe all writings off the whiteboard Clean the table by putting all objects into the basket
Figure 8 of the paper
Figure 8. Buffer length N ablation on LIBERO with chunk size C=10 fixed. (a) Success rate. (b) Per-step latency. model’s native action-chunk length (50 for π0.5). With N=5, this prefers C=10; pushing C to 15 or 20 forces the model to predict up to 100 future actions per buffer, which the short-horizon LIBERO setting does not train for. Even at C=15 and C=20, one-step async outperforms synchronous, indicating the streaming advantage persists well outside the sweet spot. Buffer length

Figures are the authors', reproduced from the paper.

References (32)
  1. [1] M. J. Kim, K. Pertsch, S. Karamcheti, T. Xiao, A. Balakrishna, S. Nair, R. Rafailov, E. Foster, G. Lam, P. 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, 2024. URL https... on PithPith's reviewarXiv
  2. [2] NVIDIA, J. Bjorck, N. C. Fernando Casta ˜neda, X. Da, R. Ding, L. J. Fan, Y . Fang, D. Fox, F. Hu, S. Huang, J. Jang, Z. Jiang, J. Kautz, K. Kundalia, L. Lao, Z. Li, Z. Lin, K. Lin, G. Liu, E. Llontop, L. Magne, A. Mandlekar, A. Narayan, S. Nasiriany, S. Reed, Y . L. Tan, G. W... on Pith
  3. [3] K. Black, N. Brown, D. Driess, A. Esmail, M. Equi, C. Finn, N. Fusai, L. Groom, K. Hausman, B. Ichter, S. Jakubczak, T. Jones, L. Ke, S. Levine, A. Li-Bell, M. Mothukuri, S. Nair, K. Pertsch, L. X. Shi, J. Tanner, Q. Vuong, A. Walling, H. Wang, and U. Zhilinsky. π0: A vision- ... on PithPith's reviewarXiv
  4. [4] W. Wu, F. Lu, Y . Wang, S. Yang, S. Liu, F. Wang, Q. Zhu, H. Sun, Y . Wang, S. Ma, Y . Ren, K. Zhang, H. Yu, J. Zhao, S. Zhou, Z. Qiu, H. Xiong, Z. Wang, Z. Wang, R. Cheng, Y .-L. Li, Y . Huang, X. Zhu, Y . Shen, and K. Zheng. A pragmatic vla foundation model, 2026. URL https:... on PithPith's reviewarXiv
  5. [5] P. Intelligence, K. Black, N. Brown, J. Darpinian, K. Dhabalia, D. Driess, A. Esmail, M. 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.... on PithPith's reviewarXiv
  6. [6] J. Tang, Y . Sun, Y . Zhao, S. Yang, Y . Lin, Z. Zhang, J. Hou, Y . Lu, Z. Liu, and S. Han. Vlash: Real-time vlas via future-state-aware asynchronous inference, 2025. URL https: //arxiv.org/abs/2512.01031 on PithPith's reviewarXiv
  7. [7] K. Black, A. Z. Ren, M. Equi, and S. Levine. Training-time action conditioning for efficient real-time chunking, 2025. URLhttps://arxiv.org/abs/2512.05964 on PitharXiv
  8. [8] Y . Shi, D. Guo, T. Zhao, F. Gao, L. Shi, C. Yu, Z. Mo, Q. Xiao, X. Peng, Q. Liao, and Y . Wang. Streamingvla: Streaming vision-language-action model with action flow matching and adaptive early observation, 2026. URLhttps://arxiv.org/abs/2603.28565 on PitharXiv
  9. [9] M. Shukor, D. Aubakirova, F. Capuano, P. Kooijmans, S. Palma, A. Zouitine, M. Aractingi, C. Pascal, M. Russi, A. Marafioti, S. Alibert, M. Cord, T. Wolf, and R. Cad`ene. Smolvla: A vision-language-action model for affordable and efficient robotics.ArXiv, abs/2506.01844, on PithPith's reviewarXiv
  10. [10] Y . Xu, Y . Yang, Z. Fan, Y . Liu, Y . Li, B. Li, and Z. Zhang. Qvla: Not all channels are equal in vision-language-action model’s quantization, 2026. URL https://arxiv.org/abs/2602. 03782 on Pith
  11. [11] Y . Yang, Y . Wang, Z. Wen, Z. Luo, C. Zou, Z. Zhang, C. Wen, and L. Zhang. Effi- cientvla: Training-free acceleration and compression for vision-language-action models.ArXiv, abs/2506.10100, 2025. URLhttps://api.semanticscholar.org/CorpusID:279318363. 11 on PithPith's reviewarXiv
  12. [12] B. Chen, D. M. Monso, Y . Du, M. Simchowitz, R. Tedrake, and V . Sitzmann. Diffusion forcing: Next-token prediction meets full-sequence diffusion, 2024. URL https://arxiv.org/abs/ 2407.01392 on PithPith's reviewarXiv
  13. [13] A. Kodaira, T. Hou, J. Hou, M. Georgopoulos, F. Juefei-Xu, M. Tomizuka, and Y . Zhao. Streamdit: Real-time streaming text-to-video generation, 2026. URL https://arxiv.org/ abs/2507.03745 on PitharXiv
  14. [14] S. ai, H. Teng, H. Jia, L. Sun, L. Li, M. Li, M. Tang, S. Han, T. Zhang, W. Q. Zhang, W. Luo, X. Kang, Y . Sun, Y . Cao, Y . Huang, Y . Lin, Y . Fang, Z. Tao, Z. Zhang, Z. Wang, Z. Liu, D. Shi, G. Su, H. Sun, H. Pan, J. Wang, J. Sheng, M. Cui, M. Hu, M. Yan, S. Yin, S. Zhang, ... on PithPith's reviewarXiv
  15. [15] J. Wen, Y . Zhu, J. Li, M. Zhu, Z. Tang, K. Wu, Z. Xu, N. Liu, R. Cheng, C. Shen, Y . Peng, F. Feng, and J. Tang. Tinyvla: Toward fast, data-efficient vision-language-action models for robotic manipulation.IEEE Robotics and Automation Letters, 10:3988–3995, 2024. URL https://a... on Pith
  16. [16] T. Lin, Y . Zhong, Y . Du, J. Zhang, J. Liu, Y . Chen, E. Gu, Z. Liu, H. Cai, Y . Zou, L. Zou, Z. Zhou, G. Li, and B. Zhao. Evo-1: Lightweight vision-language-action model with preserved semantic alignment.ArXiv, abs/2511.04555, 2025. URL https://api.semanticscholar. org/Corpu... on PitharXiv
  17. [17] H. Wang, C. Xiong, R. Wang, and X. Chen. Bitvla: 1-bit vision-language-action models for robotics manipulation.ArXiv, abs/2506.07530, 2025. URL https://api.semanticscholar. org/CorpusID:279250859 on PitharXiv
  18. [18] H. Wang, J. Xu, J. Pan, Y . Zhou, and G. Dai. Specprune-vla: Accelerating vision-language- action models via action-aware self-speculative pruning.ArXiv, abs/2509.05614, 2025. URL https://api.semanticscholar.org/CorpusID:281202916 on PithPith's reviewarXiv
  19. [19] S. Xu, Y . Wang, C. Xia, D. Zhu, T. Huang, and C. Xu. Vla-cache: Efficient vision-language- action manipulation via adaptive token caching, 2025. URL https://arxiv.org/abs/2502. 02175 on Pith
  20. [20] T. Jiang, X. Jiang, Y . Ma, X. Wen, B. Li, K. Zhan, P. Jia, Y . Liu, S. Sun, and X. Lang. The better you learn, the smarter you prune: Towards efficient vision-language-action models via differen- tiable token pruning.ArXiv, abs/2509.12594, 2025. URL https://api.semanticschola... on PitharXiv
show 12 more
  1. [21] 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, abs/2501.09747, 2025. URLhttps://api.semanticscholar.org/CorpusID:275570494 on PithPith's reviewarXiv
  2. [22] Y . Lu, Z. Liu, X. Fan, Z. Yang, J. Hou, J. Li, K. Ding, and H. Zhao. Faster: Rethinking real-time flow vlas, 2026. URLhttps://arxiv.org/abs/2603.19199 on PithPith's reviewarXiv
  3. [23] Y . Song, P. Dhariwal, M. Chen, and I. Sutskever. Consistency models. InInternational Conference on Machine Learning (ICML), 2023 on Pith
  4. [24] K. Frans, D. Hafner, S. Levine, and P. Abbeel. One step diffusion via shortcut models. In International Conference on Learning Representations (ICLR), 2025 on Pith
  5. [25] K. Black, M. Y . Galliker, and S. Levine. Real-time execution of action chunking flow policies, on Pith
  6. [26] Y . Ma, Y . Zhou, Y . Yang, T. Wang, and H. Fan. Running vlas at real-time speed, 2025. URL https://arxiv.org/abs/2510.26742. 12 on PitharXiv
  7. [27] URLhttps://arxiv.org/abs/2506.07339 on PithPith's reviewarXiv
  8. [28] B. Liu, Y . Zhu, C. Gao, Y . Feng, Q. Liu, Y . Zhu, and P. Stone. Libero: Benchmarking knowledge transfer for lifelong robot learning, 2023. URLhttps://arxiv.org/abs/2306.03310 on PithPith's reviewarXiv
  9. [29] E. Perez, F. Strub, H. de Vries, V . Dumoulin, and A. Courville. Film: Visual reasoning with a general conditioning layer, 2017. URLhttps://arxiv.org/abs/1709.07871 on PithPith's reviewarXiv
  10. [30] P. Wu, Y . Shentu, Z. Yi, X. Lin, and P. Abbeel. Gello: A general, low-cost, and intuitive teleoperation framework for robot manipulators, 2023. 13 A Appendix A.1 Simulated Experimental Settings LIBERO.We fine-tune the model on the merged training dataset from HuggingFaceVLA a... on Pith
  11. [31] T. Chen, Z. Chen, B. Chen, Z. Cai, Y . Liu, Z. Li, Q. Liang, X. Lin, Y . Ge, Z. Gu, W. Deng, Y . Guo, T. Nian, X. Xie, Q. Chen, K. Su, T. Xu, G. Liu, M. Hu, H. ang Gao, K. Wang, Z. Liang, Y . Qin, X. Yang, P. Luo, and Y . Mu. Robotwin 2.0: A scalable data generator and benchma... on PithPith's reviewarXiv
  12. [2025] URLhttps://api.semanticscholar.org/CorpusID:279119427 on Pith
Cited by 6 Pith papers
How the literature uses it

What this paper does with the arXiv papers it cites

Each citing sentence is taken from the citing paper's LaTeX source and typed by Pith's classifier. A label appears at probability 0.8 or more (0.95 for support), a contest at 0.7 or more; the quoted words are the citing authors'. How each label was checked · JSON

What it cites · 17 sentences without a confident label

Builds on 3 papers
  • Inspired by streaming chunk-wise diffusion for long video [ref] , maintains a streaming buffer of action chunks at staggered noise levels and advances them jointly under chunk-wise causal attention.

    MAGI-1: Autoregressive Video Generation at Scale · Introduction · 0.81

  • Inspired by streaming chunk-wise diffusion for long video [ref] , maintains a streaming buffer of action chunks at staggered noise levels and advances them jointly under chunk-wise causal attention.

    Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion · Introduction · 0.81

  • Inspired by streaming chunk-wise diffusion for long video [ref] , maintains a streaming buffer of action chunks at staggered noise levels and advances them jointly under chunk-wise causal attention.

    2507.03745 · Introduction · 0.81

Uses the method or tool of 2 papers
Compares with 4 papers
  • For decoding latency we compare against Realtime-VLA [ref] , the strongest system-level acceleration baseline and an instance of the per-pass-cheaper family identified in [ref].

    2510.26742 · Setup · 0.99

  • For asynchronous execution we compare against VLASH [ref] and StreamingVLA [ref] , which represent the two dominant strategies for mitigating prediction-execution mismatch (future-state conditioning and action flow matching).

    VLASH: Real-Time VLAs via Future-State-Aware Asynchronous Inference · Setup · 0.94

  • For asynchronous execution we compare against VLASH [ref] and StreamingVLA [ref] , which represent the two dominant strategies for mitigating prediction-execution mismatch (future-state conditioning and action flow matching).

    2603.28565 · Setup · 0.94

Cites as background 8 papers
  • Efficient-inference methods reduce per-step latency through lightweight architectures [ref] , compression [ref] , or token pruning [ref] , while asynchronous methods condition on future states or planned actions [ref] to suppress mismatch.

    VLASH: Real-Time VLAs via Future-State-Aware Asynchronous Inference · Introduction · 0.93

  • Efficient-inference methods reduce per-step latency through lightweight architectures [ref] , compression [ref] , or token pruning [ref] , while asynchronous methods condition on future states or planned actions [ref] to suppress mismatch.

    2512.05964 · Introduction · 0.93

  • Efficient-inference methods reduce per-step latency through lightweight architectures [ref] , compression [ref] , or token pruning [ref] , while asynchronous methods condition on future states or planned actions [ref] to suppress mismatch.

    2602.03782 · Introduction · 0.93

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment