Is In-Context Learning Learning?

Adrian de Wynter

Keywords: LLMs, in-context learning

TL;DR: A large-scale evaluation to empirically characterise in-context learning as a learning paradigm, ablating out common drawbacks of LLM evaluation, and finding results contradicting or aligning with conventional wisdom

Abstract:

In-context learning (ICL) allows some autoregressive models to solve tasks via next-token prediction and without needing further training.
This has led to claims about these model's ability to solve (learn) unseen tasks with only a few shots (exemplars) in the prompt. However, deduction does not always imply learning, as ICL does not explicitly encode a given observation. Instead, the models rely on their prior knowledge and the exemplars given, if any. We argue that, mathematically, ICL fits the definition of learning; however, its full characterisation requires empirical work. We then carry out a large-scale analysis of ICL ablating out or accounting for memorisation, pretraining, distributional shifts, and prompting style and phrasing. We find that, empirically, ICL is limited in its ability to learn and generalise to unseen tasks. Namely, in the limit where exemplars become more numerous, accuracy is insensitive to exemplar distribution, model, prompt style, and the input's linguistic features. Instead, it deduces patterns from regularities in the prompt, which leads to distributional sensitivity, especially in prompting styles such as chain-of-thought. Given the varied accuracies and on formally similar tasks, we conclude that autoregression's _ad-hoc_ encoding is not a robust mechanism for learning, and suggests limited all-purpose generalisability.

Primary Area: foundation or frontier models, including LLMs

Submission Number: 7897