Understanding In-Context Learning via Supportive Pretraining Data

TL;DR

This paper investigates the role of specific pretraining data in enabling in-context learning (ICL) in language models, revealing that targeted data supports ICL and can significantly improve model performance.

Contribution

It introduces an iterative, gradient-based method to identify pretraining data that supports ICL and demonstrates that training on this data enhances ICL ability.

Findings

Supportive pretraining data do not have higher domain relevance.

Supportive data contain more long-tail, rare tokens.

Supportive data are challenging examples with low information gain from long-range context.

Abstract

In-context learning (ICL) improves language models' performance on a variety of NLP tasks by simply demonstrating a handful of examples at inference time. It is not well understood why ICL ability emerges, as the model has never been specifically trained on such demonstrations. Unlike prior work that explores implicit mechanisms behind ICL, we study ICL via investigating the pretraining data. Specifically, we first adapt an iterative, gradient-based approach to find a small subset of pretraining data that supports ICL. We observe that a continued pretraining on this small subset significantly improves the model's ICL ability, by up to 18%. We then compare the supportive subset constrastively with random subsets of pretraining data and discover: (1) The supportive pretraining data to ICL do not have a higher domain relevance to downstream tasks. (2) The supportive pretraining data have a higher mass of rarely occurring, long-tail tokens. (3) The supportive pretraining data are challenging examples where the information gain from long-range context is below average, indicating learning to incorporate difficult long-range context encourages ICL. Our work takes a first step towards understanding ICL via analyzing instance-level pretraining data. Our insights have a potential to enhance the ICL ability of language models by actively guiding the construction of pretraining data in the future.