HYPE: Hyperbolic Entailment Filtering for Underspecified Images and Texts

TL;DR

HYPE introduces hyperbolic entailment filtering to improve data quality in image-text datasets, leading to better self-supervised learning by selecting more meaningful and specific samples, and achieving state-of-the-art results.

Contribution

The paper presents a novel hyperbolic embedding-based filtering method that enhances data specificity and alignment in large-scale datasets for self-supervised learning.

Findings

HYPE significantly improves filtering efficiency.

Sets new state-of-the-art on DataComp benchmark.

Image specificity measure outperforms CLIP score for dataset induction.

Abstract

In an era where the volume of data drives the effectiveness of self-supervised learning, the specificity and clarity of data semantics play a crucial role in model training. Addressing this, we introduce HYPerbolic Entailment filtering (HYPE), a novel methodology designed to meticulously extract modality-wise meaningful and well-aligned data from extensive, noisy image-text pair datasets. Our approach leverages hyperbolic embeddings and the concept of entailment cones to evaluate and filter out samples with meaningless or underspecified semantics, focusing on enhancing the specificity of each data sample. HYPE not only demonstrates a significant improvement in filtering efficiency but also sets a new state-of-the-art in the DataComp benchmark when combined with existing filtering techniques. This breakthrough showcases the potential of HYPE to refine the data selection process, thereby contributing to the development of more accurate and efficient self-supervised learning models. Additionally, the image specificity $\epsilon_{i}$ can be independently applied to induce an image-only dataset from an image-text or image-only data pool for training image-only self-supervised models and showed superior performance when compared to the dataset induced by CLIP score.