This paper focuses on improving task embedding for meta/few-shot learning. The proposal is a model-aware task embedding that can take the task difficulty into account. The philosophy behind sounds quite interesting to me, namely, the model trained for each task is itself a key property of the task when considering transferring knowledge between different tasks. This philosophy leads to a novel algorithm design I have never seen. The clarity and novelty are clearly above the bar of NeurIPS. While the reviewers had some concerns on the significance, the authors did a particularly good job in their rebuttal. Thus, all of us have agreed to accept this paper for publication! Please include the additional experimental results in the next version.