Meta-learning a transformer on a curated collection (Meta-Album) instead of a single large dataset (ImageNet-1k) maintains or improves few-shot in-context generalization, and sequential presentation can further improve it.
A pre-processing step is applied to remove special characters and convert all names to lowercase, ensuring consistency in the comparison
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Meta-Learning Transformers to Improve In-Context Generalization
Meta-learning a transformer on a curated collection (Meta-Album) instead of a single large dataset (ImageNet-1k) maintains or improves few-shot in-context generalization, and sequential presentation can further improve it.