Decompose and Recompose: Reasoning New Skills from Existing Abilities for Cross-Task Robotic Manipulation

TL;DR

This paper introduces a framework that decomposes demonstrations into interpretable skills, enabling robotic systems to recombine and generalize skills for unseen tasks without retraining.

Contribution

It proposes a novel skill reasoning approach using atomic skill-action pairs, improving cross-task generalization in robotic manipulation.

Findings

Achieves zero-shot cross-task generalization on AGNOSTOS benchmark.

Effectively recomposes skills for unseen tasks in real-world environments.

Outperforms existing methods by capturing composable skill knowledge.

Abstract

Cross-task generalization is a core challenge in open-world robotic manipulation, and the key lies in extracting transferable manipulation knowledge from seen tasks. Recent in-context learning approaches leverage seen task demonstrations to generate actions for unseen tasks without parameter updates. However, existing methods provide only low-level continuous action sequences as context, failing to capture composable skill knowledge and causing models to degenerate into superficial trajectory imitation. We propose Decompose and Recompose, a skill reasoning framework using atomic skill-action pairs as intermediate representations. Our approach decomposes seen demonstrations into interpretable skill--action alignments, enabling the model to recompose these skills for unseen tasks through compositional reasoning. Specifically, we construct a task-adaptive dynamic demonstration library via visual-semantic retrieval combined with skill sequences from a planning agent, complemented by a coverage-aware static library to fill missing skill patterns. Together, these yield skill-comprehensive demonstrations that explicitly elicit compositional reasoning for skill composition and execution ordering. Experiments on the AGNOSTOS benchmark and real-world environments validate our method's zero-shot cross-task generalization capability.