Latent Bridge: Feature Delta Prediction for Efficient Dual-System Vision-Language-Action Model Inference
Yudong Liu, Yuan Li, Zijia Tang, Yuxi Zheng, Yueqian Lin + 10 more
TLDR
Latent Bridge predicts VLM feature deltas to enable efficient dual-system VLA model inference, significantly speeding up robotic manipulation.
Key contributions
- Predicts VLM output deltas, allowing periodic VLM calls for efficient VLA inference.
- Generalizes across distinct VLA architectures (e.g., GR00T-N1.6, π0.5).
- Uses a task-agnostic DAgger training pipeline for broad applicability.
- Retains 95-100% performance while reducing VLM calls by 50-75%, speeding up inference 1.65-1.73x.
Why it matters
Dual-system VLA models are powerful but slow due to frequent VLM calls. Latent Bridge addresses this bottleneck by predicting feature deltas, enabling significant speedups without sacrificing performance. This makes advanced robotic manipulation more practical and deployable.
Original Abstract
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.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.