Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

General-Purpose In-Context Learning by Meta-Learning Transformers

About

Modern machine learning requires system designers to specify aspects of the learning pipeline, such as losses, architectures, and optimizers. Meta-learning, or learning-to-learn, instead aims to learn those aspects, and promises to unlock greater capabilities with less manual effort. One particularly ambitious goal of meta-learning is to train general-purpose in-context learning algorithms from scratch, using only black-box models with minimal inductive bias. Such a model takes in training data, and produces test-set predictions across a wide range of problems, without any explicit definition of an inference model, training loss, or optimization algorithm. In this paper we show that Transformers and other black-box models can be meta-trained to act as general-purpose in-context learners. We characterize transitions between algorithms that generalize, algorithms that memorize, and algorithms that fail to meta-train at all, induced by changes in model size, number of tasks, and meta-optimization. We further show that the capabilities of meta-trained algorithms are bottlenecked by the accessible state size (memory) determining the next prediction, unlike standard models which are thought to be bottlenecked by parameter count. Finally, we propose practical interventions such as biasing the training distribution that improve the meta-training and meta-generalization of general-purpose in-context learning algorithms.

Louis Kirsch, James Harrison, Jascha Sohl-Dickstein, Luke Metz• 2022

Related benchmarks

TaskDatasetResultRank
Few-shot classificationCUB (test)
Accuracy94.5
145
Few-shot Image ClassificationminiImageNet (test)--
111
Few-shot Image ClassificationtieredImageNet (test)
Accuracy97.2
86
Few-shot classificationChestX (test)
Accuracy20.9
46
Few-shot classificationmeta-iNat (test)
Accuracy95.1
34
Few-shot classificationtiered meta-iNat (test)
Accuracy87.6
34
Few-shot Image ClassificationAircraft (test)
Mean Accuracy61.8
28
Showing 7 of 7 rows

Other info

Follow for update