When LLMs are Unfit Use FastFit: Fast and Effective Text Classification with Many Classes

arXiv:2404.12365v1 Announce Type: cross
Abstract: We present FastFit, a method, and a Python package design to provide fast and accurate few-shot classification, especially for scenarios with many semantically similar classes. FastFit utilizes a novel approach integrating batch contrastive learning and token-level similarity score. Compared to existing few-shot learning packages, such as SetFit, Transformers, or few-shot prompting of large language models via API calls, FastFit significantly improves multiclass classification performance in speed and accuracy across FewMany, our newly curated English benchmark, …

abstract arxiv classification cs.ai cs.cl cs.ir cs.lg design few-shot llms novel package python text text classification token type