mlsanityData to 7 Oct 2026

Latent Bridge: Feature Delta Prediction for Efficient Dual-System Vision-Language-Action Model Inference

arXiv 2605.02739 ↗First posted 4 May 2026cs.ROcited by 4

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

Abstract (arXiv)

Dual-system Vision-Language-Action (VLA) models achieve state-of-the-art robotic manipulation but are bottlenecked by the VLM backbone, which must execute at every control step while producing temporally redundant features. We propose Latent Bridge, a lightweight model that predicts VLM output deltas between timesteps, enabling the action head to operate on predicted outputs while the expensive VLM backbone is called only periodically. We instantiate Latent Bridge on two architecturally distinct VLAs: GR00T-N1.6 (feature-space bridge) and π0.5 (KV-cache bridge), demonstrating that the approach generalizes across VLA designs. Our task-agnostic DAgger training pipeline transfers across benchmarks without modification. Across four LIBERO suites, 24 RoboCasa kitchen tasks, and the ALOHA sim transfer-cube task, Latent Bridge achieves 95-100% performance retention while reducing VLM calls by 50-75%, yielding 1.65-1.73x net per-episode speedup.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Placed in Efficient VLA inference and adaptation, one of 13 directions in Vision-language-action models. Builds on 30 of its 40 references that are in the map. First posted on arXiv 4 May 2026.

Builds on

30 of 40 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Compressor-VLA: Instruction-Guided Visual Token Compression for Efficient Robotic Manipulation24 Nov
  2. 2025The Better You Learn, The Smarter You Prune: Towards Efficient Vision-language-action Models via Differentiable Token Pruning16 Sep
  3. 2025VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching4 Feb

Other references · 27

  1. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  2. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  3. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,730
  4. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  5. 2023Efficient Streaming Language Models with Attention Sinkscited by 2,115
  6. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
  7. 2022GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformerscited by 1,838
  8. 2023Consistency Modelscited by 1,778
  9. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  10. 2022Fast Inference from Transformers via Speculative Decodingcited by 1,615
  11. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,598
  12. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
15 more
  1. 2024Octo: An Open-Source Generalist Robot Policycited by 1,427
  2. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,338
  3. 2023Mastering Diverse Domains through World Modelscited by 1,193
  4. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,167
  5. 2022Token Merging: Your ViT But Fastercited by 995
  6. 2024SnapKV: LLM Knows What You are Looking for Before Generationcited by 848
  7. 2024Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Headscited by 821
  8. 2018Recurrent World Models Facilitate Policy Evolutioncited by 761
  9. 2024Genie: Generative Interactive Environmentscited by 706
  10. 2024An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Modelscited by 629
  11. 2024Revisiting Feature Prediction for Learning Visual Representations from Videocited by 490
  12. 2023Learning Interactive Real-World Simulatorscited by 440
  13. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 411
  14. 2024RoboCasa: Large-Scale Simulation of Everyday Tasks for Generalist Robotscited by 391
  15. 2024Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformerscited by 145

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 4.

4 papers.

  1. 2026Jetson-PI: Towards Onboard Real-Time Robot Control via Foresight-Aligned Asynchronous Inferencecited by 3
  2. 2026Enfold: Folding World Model Imagination into Predictive Representations for Ultra-Efficient Embodied Controlcited by 3
  3. 2026PonderPounce: A Pretrained MLLM as an Episode Context Engine for Robot Controlcited by 0
  4. 2026VLA-ULAP: Interleaving Cloud VLA Calls with Ultra-Lightweight Local Action Prediction at the Edgecited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Finetuning Vision-Language-Action Models Requires Fewer Layers Than You Think
  2. 2026A1: A Fully Transparent Open-Source, Adaptive and Efficient Truncated Vision-Language-Action Model
  3. 2026Co-VLA: Coordination-Aware Structured Action Modeling for Dual-Arm Vision-Language-Action Systems
  4. 2025Think Twice, Act Once: Token-Aware Compression and Action Reuse for Efficient Inference in Vision-Language-Action Models
  5. 2025FASTer: Toward Efficient Autoregressive Vision Language Action Modeling via Neural Action Tokenization
  6. 2026ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning
  7. 2026Reducing Temporal Redundancy for Efficient Vision-Language-Action Inference
  8. 2026CLAP: Direct VLM-to-VLA Adaptation via Language-Action Grounding