mlsanityData to 14 Oct 2026

Realtime-VLA FLASH: Speculative Inference Framework for Diffusion-based VLAs

arXiv 2605.13778 ↗First posted 13 May 2026cs.RO cs.CVcited by 8

On the map: Vision-language-action models → Efficient VLA inference and adaptation

also in 1 more
Abstract (arXiv)

Diffusion-based vision-language-action models (dVLAs) are promising for embodied intelligence but are fundamentally limited in real-time deployment by the high latency of full inference. We propose Realtime-VLA FLASH, a speculative inference framework that eliminates most full inference calls during replanning by introducing a lightweight draft model with parallel verification via the main model's Action Expert and a phase-aware fallback mechanism that reverts to the full inference pipeline when needed. This design enables low-latency, high-frequency replanning without sacrificing reliability. Experiments show that on LIBERO, FLASH largely preserves task performance by replacing many 58.0 ms full-inference rounds with speculative rounds as fast as 7.8 ms, lowering task-level average inference latency to 19.1 ms (3.04x speedup). We additionally demonstrate effectiveness on real-world conveyor-belt sorting, highlighting its practical impact for latency-critical embodied tasks.

Key facts. Cited by 8 papers in the map (data through 14 Oct 2026). Placed in Efficient VLA inference and adaptation, one of 13 directions in Vision-language-action models. Builds on 31 of its 32 references that are in the map. First posted on arXiv 13 May 2026.

Builds on

31 of 32 references.

Closest ancestors · 8

Same direction, newest first

  1. 2026KERV: Kinematic-Rectified Speculative Decoding for Embodied VLA Models2 Mar
  2. 2026Mean-Flow based One-Step Vision-Language-Action2 Mar
  3. 2026QuantVLA: Scale-Calibrated Post-Training Quantization for Vision-Language-Action Models23 Feb
  4. 2026How Fast Can I Run My VLA? Demystifying VLA Inference Performance with VLA-Perf20 Feb
  5. 2025Training-Time Action Conditioning for Efficient Real-Time Chunking5 Dec
  6. 2025SpecPrune-VLA: Accelerating Vision-Language-Action Models via Action-Aware Self-Speculative Pruning6 Sep
  7. 2025Spec-VLA: Speculative Decoding for Vision-Language-Action Models with Relaxed Acceptance30 Jul
  8. 2025EfficientVLA: Training-Free Acceleration and Compression for Vision-Language-Action Models11 Jun

Other references · 23

  1. 2022Flow Matching for Generative Modelingcited by 5,702
  2. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,838
  3. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
  4. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,998
  5. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,568
  6. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,410
  7. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,225
  8. 2024EAGLE: Speculative Sampling Requires Rethinking Feature Uncertaintycited by 578
  9. 2025Mean Flows for One-step Generative Modelingcited by 561
  10. 2025SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Roboticscited by 530
  11. 2024One Step Diffusion via Shortcut Modelscited by 391
  12. 2024EAGLE-2: Faster Inference of Language Models with Dynamic Draft Treescited by 370
11 more
  1. 2024TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulationcited by 339
  2. 2025EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Testcited by 331
  3. 2025Real-Time Execution of Action Chunking Flow Policiescited by 216
  4. 2024One-Step Diffusion Policy: Fast Visuomotor Policies via Diffusion Distillationcited by 89
  5. 2025Running VLAs at Real-time Speedcited by 43
  6. 2025Evo-1: Lightweight Vision-Language-Action Model with Preserved Semantic Alignmentcited by 41
  7. 2025Accelerated Diffusion Models via Speculative Samplingcited by 31
  8. 2026RDT2: Exploring the Scaling Limit of UMI Data Towards Zero-Shot Cross-Embodiment Generalizationcited by 30
  9. 2025Diffusion Models are Secretly Exchangeable: Parallelizing DDPMs via Autospeculationcited by 11
  10. 2025Lite VLA: Efficient Vision-Language-Action Control on CPU-Bound Edge Robotscited by 9
  11. 2026Realtime-VLA V2: Learning to Run VLAs Fast, Smooth, and Accuratecited by 8

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 8.

8 papers.

  1. 2026TurboVLA: Real-Time Vision-Language-Action Model at 32 Hz on an RTX 4090 with <1 GB VRAMcited by 10
  2. 2026Jetson-PI: Towards Onboard Real-Time Robot Control via Foresight-Aligned Asynchronous Inferencecited by 3
  3. 2026Reinforcement Learning for Real-Time Vision-Language-Action Policiescited by 2
  4. 2026Is Multimodal Speculative Decoding Ready for Diffusion-Based Parallel Drafting? A Survey and Empirical Diagnosiscited by 0
  5. 2026VLA-ULAP: Interleaving Cloud VLA Calls with Ultra-Lightweight Local Action Prediction at the Edgecited by 0
  6. 2026rMuscle: Robotic Muscle Memory for Efficient Vision-Language-Action Model Inferencecited by 0
  7. 2026Faster and Better? Benchmark Bugs and Design Limitations Distort the Evaluation of Vision-Language-Action Accelerationcited by 0
  8. 2026Toward Real-Time VLAs: Stage-Aware Two-Step Flow Denoising and System-Level Evaluationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Deltoris: Enabling Real-time VLA Inference in Embodied AI via Bit-level Sparsity and Speculative Inference
  2. 2026FlashVLA: Streaming Action Decoding for Fast and Asynchronous VLA Inference
  3. 2026Algorithm-Architecture Co-Design for Efficient VLA Inference via Speculative Inference and Verification
  4. 2026AC^2-VLA: Action-Context-Aware Adaptive Computation in Vision-Language-Action Models for Efficient Robotic Manipulation
  5. 2026Reducing Temporal Redundancy for Efficient Vision-Language-Action Inference
  6. 2026A1: A Fully Transparent Open-Source, Adaptive and Efficient Truncated Vision-Language-Action Model
  7. 2026Recurrent-Depth VLA: Implicit Test-Time Compute Scaling of Vision-Language-Action Models via Latent Iterative Reasoning
  8. 2026Beyond LLM Serving: Characterizing Vision-Language-Action Workloads for Embodied AI System Design