CoLLEGe: Concept Embedding Generation for Large Language Models

TL;DR

CoLLEGe introduces a meta-learning framework that enables large language models to quickly generate embeddings for new concepts using few examples, improving on existing methods for concept learning without task-specific training.

Contribution

The paper presents CoLLEGe, a novel meta-learning approach for few-shot concept embedding generation tailored for large language models, enhancing their ability to learn new concepts efficiently.

Findings

Successfully learns new word concepts without task-specific training

Effective in definition inference and verbal reasoning tasks

Outperforms baseline methods in real-world scenarios

Abstract

Current language models are unable to quickly learn new concepts on the fly, often requiring a more involved finetuning process to learn robustly. Prompting in-context is not robust to context distractions, and often fails to confer much information about the new concepts. Classic methods for few-shot word learning in NLP, relying on global word vectors, are less applicable to large language models. In this paper, we introduce a novel approach named CoLLEGe (Concept Learning with Language Embedding Generation) to modernize few-shot concept learning. CoLLEGe is a meta-learning framework capable of generating flexible embeddings for new concepts using a small number of example sentences or definitions. Our primary meta-learning objective is simply to facilitate a language model to make next word predictions in forthcoming sentences, making it compatible with language model pretraining. We design a series of tasks to test new concept learning in challenging real-world scenarios, including new word acquisition, definition inference, and verbal reasoning, and demonstrate that our method succeeds in each setting without task-specific training. Code and data for our project can be found at https://college-concept-learning.github.io/