ElasticFlow: One-Step Physics-Consistent Policy with Elastic Time Horizons for Language-Guided Manipulation

2026-05-09

AI Summary

This work addresses the high inference latency of diffusion policies in embodied intelligence, which often compromises physical consistency when accelerated by existing methods. The authors propose a distillation-free, single-step policy framework that directly models the mean velocity field grounded in mean-field theory and introduces an elastic time-horizon mechanism to explicitly align the control granularity of language instructions with physical execution. This approach effectively handles task temporal heterogeneity and spectral bias while preserving physical plausibility. To the best of our knowledge, it is the first method to achieve physically consistent single-step inference, attaining real-time performance of approximately 71 Hz (1 NFE) on the LIBERO, CALVIN, and RoboTwin benchmarks and outperforming state-of-the-art models such as OpenVLA and π₀ on long-horizon tasks.

Abstract

Diffusion policies have demonstrated exceptional performance in embodied AI. However, their iterative denoising process results in high latency, and existing acceleration methods often sacrifice physical consistency. To address this, we propose ElasticFlow, a distillation-free, physics-consistent one-step policy framework. We reconstruct the Mean Field Theory by directly modeling the average velocity field, enabling a direct single-step mapping from noise to action. Addressing the Temporal Heterogeneity of robotic tasks, we introduce the Elastic Time Horizons mechanism. This mechanism effectively overcomes Spectral Bias by explicitly encoding control granularity, achieving efficient alignment between semantic instructions and physical execution horizons. Experiments on benchmarks such as LIBERO, CALVIN, and RoboTwin demonstrate that ElasticFlow achieves efficient 1-NFE inference (approximately 71Hz). Furthermore, it outperforms state-of-the-art methods, including OpenVLA and $π_0$, on long-horizon tasks, highlighting its potential for efficient, robust, and semantically aligned control.

Problem

• diffusion policies
• physical consistency
• temporal heterogeneity
• semantic alignment
• latency

Innovation

• ElasticFlow
• one-step policy
• physics consistency
• elastic time horizons
• mean field theory