Has Your Pretrained Model Improved? A Multi-head Posterior Based Approach
This addresses the need for better evaluation methods in NLP and computer vision, offering a more efficient alternative to fine-tuning for assessing model improvements.
The paper tackles the problem of efficiently evaluating pretrained models by proposing a novel metric based on the consistency between entity representations and meta-features, demonstrating effectiveness across domains like relational datasets, large language models, and image models.
The emergence of pre-trained models has significantly impacted Natural Language Processing (NLP) and Computer Vision to relational datasets. Traditionally, these models are assessed through fine-tuned downstream tasks. However, this raises the question of how to evaluate these models more efficiently and more effectively. In this study, we explore a novel approach where we leverage the meta-features associated with each entity as a source of worldly knowledge and employ entity representations from the models. We propose using the consistency between these representations and the meta-features as a metric for evaluating pre-trained models. Our method's effectiveness is demonstrated across various domains, including models with relational datasets, large language models and image models.