Problem
Research questions and friction points this paper is trying to address.
In-context imitation learning
Few-shot policy generalization
Cross-object transfer
Unseen objects
MatchingPolicy: Correspondence-Aware Policy Enables Cross-Object In-Context Learning
🤖 AI Summary
This study addresses the degradation of generalization performance in context-based imitation learning when applied to unseen objects and novel environments. To overcome this limitation, we propose a Correspondence-Aware Diffusion Policy that explicitly decouples demonstration matching from policy learning. By integrating vision foundation models with a two-stage matching algorithm, our method establishes a robust cross-object few-shot transfer mechanism. Experimental evaluations on both the RLBench benchmark and real-world robotic manipulation tasks demonstrate that this architecture achieves superior few-shot adaptability and cross-category generalization. Consequently, the proposed approach significantly enhances deployment efficiency and stability in new environments, effectively mitigating the domain gap inherent in traditional context-dependent imitation learning frameworks.
📝 Abstract
In-context imitation learning enables few-shot policy generalization but struggles to maintain performance on unseen objects and novel scenarios. To address this, we introduce MatchingPolicy, a correspondence-driven framework that explicitly decouples demonstration-to-scene matching from policy learning. Central to our method is a correspondence-aware diffusion policy that conditions robotic actions directly on dense semantic correspondences. This architectural separation resolves the inherent conflict between correspondence identification and action adaptation, enabling robust out-of-distribution transfer. Our framework integrates vision foundation models with a novel two-stage matching algorithm to dynamically establish reliable correspondences. Extensive evaluations on RLBench and real-world manipulation tasks confirm that MatchingPolicy achieves superior few-shot performance, generalizing reliably across unseen object instances and semantic categories.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Correspondence-Aware Diffusion Policy
Cross-Object In-Context Learning
Decoupled Matching and Policy Learning
Vision Foundation Models
Few-Shot Generalization
🔎 Similar Papers
No similar papers found.
