World In-Context Learning Revolutionizes Robot Manipulation

News

Researchers introduce World In-Context Learning (WICL) and RoboScript, achieving 83.4% success in robotic tasks with a single demonstration.

Researchers from Huazhong University of Science and Technology and ArcheOne AI have unveiled a groundbreaking approach to robotic manipulation called World In-Context Learning (WICL). Published on November 11, 2026, on arXiv, the study introduces RoboScript, an open-source framework that achieves state-of-the-art performance in robotic tasks using a single demonstration without gradient updates.

World In-Context Learning: A Paradigm Shift

Traditional trained action policies encode manipulation skills into network weights, limiting their adaptability to new environments. In contrast, WICL decouples learning from weights by leveraging in-context learning. A frozen foundation model starts from a single demonstration, rehearses the task in a simulated environment, and stores the learned knowledge as a Task Memory. This memory is then deployed in new, unseen scenes, ensuring adaptability across diverse visual distributions.

Performance and Key Results

RoboScript was validated on RoboTwin 2.0, a benchmark of 50 bimanual manipulation tasks. The framework achieved an 83.4% mean success rate across clean and randomized settings, outperforming the strongest trained action policy by over 20 percentage points. Notably, this success was achieved with just one demonstration per task, compared to 50 demonstrations required by traditional methods. The system’s performance remained consistent across randomized scenes, with success rates dropping minimally from 83.6% to 83.2%.

Technical Innovations

RoboScript’s modular infrastructure enables robots to learn tasks in context rather than weights. Key features include:

• Task Memory: Knowledge is stored in context, allowing for seamless transfer to new environments.
• Rehearsal in Simulation: Tasks are practiced in a simulated world to refine understanding without altering model weights.
• Cross-Embodiment Adaptation: The framework successfully transfers learned tasks to physical dual-arm robots, demonstrating real-world applicability.

Open-Source Availability

The RoboScript framework, along with its implementation of WICL, is available under the Apache License 2.0 on GitHub. The repository, created on November 5, 2026, has already garnered community interest, with 34 stars and no open issues as of the publication date. Researchers and developers can access the code at https://github.com/ArcheOne-AI/RoboScript.

Implications and Future Directions

WICL represents a significant advancement in robotic manipulation, enabling robots to learn tasks with minimal data and adapt to new environments without retraining. The approach’s scalability with foundation models and its ability to improve through reflection suggest broad potential applications in industrial automation, service robotics, and beyond. Further validation on additional benchmarks and real-world scenarios is anticipated to solidify its impact.