In-Context Learning for Robots: Methods and Applications

arXiv:2609.36012v1 Announce Type: new Abstract: General-purpose robots must infer what a new task requires and translate that understanding into appropriate physical action. In-context learning (ICL) for robots supports this process by using demonstrations and interaction to direct existing competence with neural parameters held fixed during deployment. We organize this literature review around the interfaces connecting contextual evidence to execution, distinguishing four families: context-conditioned policies, geometric demonstration transfer, world-model-based control, and skill- and agent-based execution. Comparing these interfaces clarifies their transfer assumptions and the roles of training, correspondence, and memory in making context useful. Across manipulation and navigation, we examine how these mechanisms preserve taught requirements as objects, environments, and execution conditions change. This analysis links method design to evaluation practices that distinguish responsiveness to teaching, physical transfer, and benefits from retained experience. The resulting agenda connects compositional task acquisition and faithful transfer with physical recursive self-improvement, in which experience improves the ability to learn subsequent tasks.

Abstract:General-purpose robots must infer what a new task requires and translate that understanding into appropriate physical action. In-context learning (ICL) for robots supports this process by using demonstrations and interaction to direct existing competence with neural parameters held fixed during deployment. We organize this literature review around the interfaces connecting contextual evidence to execution, distinguishing four families: context-conditioned policies, geometric demonstration transfer, world-model-based control, and skill- and agent-based execution. Comparing these interfaces clarifies their transfer assumptions and the roles of training, correspondence, and memory in making context useful. Across manipulation and navigation, we examine how these mechanisms preserve taught requirements as objects, environments, and execution conditions change. This analysis links method design to evaluation practices that distinguish responsiveness to teaching, physical transfer, and benefits from retained experience. The resulting agenda connects compositional task acquisition and faithful transfer with physical recursive self-improvement, in which experience improves the ability to learn subsequent tasks.

100 pages, 26 figures, 25 tables. Project page: this https URL ; Code and literature: this https URL

Robotics (cs.RO); Machine Learning (cs.LG)

arXiv:2609.36012 [cs.RO]

(or arXiv:2609.36012v1 [cs.RO] for this version)

https://doi.org/10.48550/arXiv.2609.36012 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

Supplement-S1/CODEBOOK.md

Supplement-S1/README.md

Supplement-S1/REPORTED-COMPARISONS.md

Supplement-S1/annual-counts.csv

Supplement-S1/checksums.sha256

Supplement-S1/codebook.json

Supplement-S1/corpus.csv

Supplement-S1/corpus.json

Supplement-S1/coverage-counts.csv

Supplement-S1/family-counts.csv

Supplement-S1/references.csv

Supplement-S1/references.json

Supplement-S1/reported-comparisons.csv

Supplement-S1/reported-comparisons.json

Supplement-S1/reproduce.py

Supplement-S1/role-counts.csv

Supplement-S1/summary.json

(12 additional files not shown) You must enabled JavaScript to view entire file list.

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

aiqbench analysis

In-Context Learning for Robots: Methods and Applications

Related coverage