Training-Free Imitation Learning with Closed-Form Diffusion Policies

2026-05-31

AI Summary

Traditional diffusion-based policies rely on time-consuming offline training, which impedes rapid iteration in data collection and deployment. This work proposes Closed-Form Diffusion Policies (CFDP), the first training-free diffusion policy framework that derives a closed-form score function directly from demonstration data, enabling real-time inference on CPU at millisecond latency. The method matches or exceeds the performance of neural baselines requiring hours of training while serving as a composable primitive that allows flexible, on-the-fly editing of pretrained policies during inference. By eliminating the need for iterative optimization, CFDP significantly enhances both the efficiency and adaptability of imitation learning systems.

Abstract

While diffusion-based policies have impressive performance and expressivity, their long offline training slows down the data collection and policy deployment loop. We introduce Closed-Form Diffusion Policies, a class of training-free diffusion-based policies for imitation learning using the closed-form score derived from the demonstration dataset. We deploy CFDP with real-time inference with a mobile CPU in hardware experiments, showing it can successfully perform imitation directly from the dataset in milliseconds and with faster inference than neural diffusion policies. In experiments on imitation learning benchmarks, we show that CFDP is competitive against neural baselines that require hours of training, providing a favorable tradeoff between training time and performance. Finally, we show how closed-form diffusion policies act as a composable primitive that enables data-driven inference-time editing of pre-trained neural diffusion policies, including policy guidance and novel demonstration augmentation.

Problem

• Imitation Learning
• Diffusion Policies
• Training-Free
• Policy Deployment
• Demonstration Dataset

Innovation

• training-free
• closed-form diffusion
• imitation learning
• real-time inference
• policy editing