# The effects of negative adaptation in Model-Agnostic Meta-Learning

arXiv: https://arxiv.org/abs/1812.02159 · HTML: https://mlsanity.com/paper/1812.02159/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 5 Dec 2018.

## Built on by (most cited first, 6 of 6)

- [Few-Shot Class-Incremental Learning by Sampling Multi-Phase Tasks](https://mlsanity.com/paper/2203.17030/) (2022)
- [Transfer Learning in Robotics: An Upcoming Breakthrough? A Review of Promises and Challenges](https://mlsanity.com/paper/2311.18044/) (2023)
- [Distributionally Adaptive Meta Reinforcement Learning](https://mlsanity.com/paper/2210.03104/) (2022)
- [What Can Knowledge Bring to Machine Learning? -- A Survey of Low-shot Learning for Structured Data](https://mlsanity.com/paper/2106.06410/) (2021)
- [Distribution Embedding Networks for Generalization from a Diverse Set of Classification Tasks](https://mlsanity.com/paper/2202.01940/) (2022)
- [When Does Adaptation Win? Scaling Laws for Meta-Learning in Quantum Control](https://mlsanity.com/paper/2601.18973/) (2026)

## Abstract (arXiv)

The capacity of meta-learning algorithms to quickly adapt to a variety of tasks, including ones they did not experience during meta-training, has been a key factor in the recent success of these methods on few-shot learning problems. This particular advantage of using meta-learning over standard supervised or reinforcement learning is only well founded under the assumption that the adaptation phase does improve the performance of our model on the task of interest. However, in the classical framework of meta-learning, this constraint is only mildly enforced, if not at all, and we only see an improvement on average over a distribution of tasks. In this paper, we show that the adaptation in an algorithm like MAML can significantly decrease the performance of an agent in a meta-reinforcement learning setting, even on a range of meta-training tasks.