On the Compositional Generalization Gap of In-Context Learning
This addresses the problem of compositional generalization in large language models for NLP researchers, showing incremental improvements with scaling.
The study investigated the compositional generalization gap between in-distribution and out-of-distribution performance in large language models using in-context learning for semantic parsing tasks, finding that the relative generalization gap decreases as model size scales up across four model families and three datasets.
Pretrained large generative language models have shown great performance on many tasks, but exhibit low compositional generalization abilities. Scaling such models has been shown to improve their performance on various NLP tasks even just by conditioning them on a few examples to solve the task without any fine-tuning (also known as in-context learning). In this work, we look at the gap between the in-distribution (ID) and out-of-distribution (OOD) performance of such models in semantic parsing tasks with in-context learning. In the ID settings, the demonstrations are from the same split (test or train) that the model is being evaluated on, and in the OOD settings, they are from the other split. We look at how the relative generalization gap of in-context learning evolves as models are scaled up. We evaluate four model families, OPT, BLOOM, CodeGen and Codex on three semantic parsing datasets, CFQ, SCAN and GeoQuery with different number of exemplars, and observe a trend of decreasing relative generalization gap as models are scaled up.