SeedPolicy: Horizon Scaling via Self-Evolving Diffusion Policy for Robot Manipulation
For robotic imitation learning, SeedPolicy addresses the temporal modeling bottleneck in Diffusion Policy, enabling more effective long-horizon manipulation with high efficiency.
SeedPolicy introduces a Self-Evolving Gated Attention module to extend the effective temporal horizon of Diffusion Policy for robot manipulation, achieving 36.8% relative improvement in clean settings and 169% in randomized challenging settings on the RoboTwin 2.0 benchmark, outperforming larger models like RDT.
Imitation Learning (IL) enables robots to acquire manipulation skills from expert demonstrations. Diffusion Policy (DP) models multi-modal expert behaviors but degrades when naively increasing stacked observation horizons, limiting long-horizon manipulation. We propose Self-Evolving Gated Attention (SEGA), a temporal module that maintains a time-evolving latent state via gated attention, enabling efficient recurrent updates that accumulate long-term context into a compact latent representation while filtering irrelevant temporal information. Integrating SEGA into DP yields Self-Evolving Diffusion Policy (SeedPolicy), which resolves the temporal modeling bottleneck and extends the effective temporal horizon with moderate overhead. On the RoboTwin 2.0 benchmark with 50 manipulation tasks, SeedPolicy outperforms DP and other IL baselines. Averaged across both CNN and Transformer backbones, SeedPolicy achieves 36.8% relative improvement in clean settings and 169% relative improvement in randomized challenging settings over the DP. Compared to vision-language-action models such as RDT with 1.2B parameters, SeedPolicy achieves stronger performance in the clean setting with one to two orders of magnitude fewer parameters, demonstrating strong efficiency. These results establish SeedPolicy as a state-of-the-art imitation learning method for long-horizon robotic manipulation. Code is available at: https://anonymous.4open.science/r/SeedPolicy-64F0/.