A major challenge in robotics is to deploy simulated controllers into real-world.

Moreover, the training and storage costs of such method grow linearly with the number of potential scenarios.

Object, attribute or stylistic changes can be learned from visually static datasets. On the other hand, high-quality data for action and reasoning-centric edits is scarce and has to come from entirely different sources that cover e.g.

To bridge this disparity, some works fine-tune LLMs to relearn reasoning capabilities in non-English languages, while others replace non-English inputs with an external model's outputs

For each prediction task, we select features that have a causal influence on the target of prediction.