How Do Transformers Learn In-Context Beyond Simple Functions? A Case Study on Learning with Representations

Open in new window