Structure Retention in Embedding Spaces as a Predictor of Benchmark Performance

TL;DR

This paper demonstrates that the organization of embedding spaces in high-performing models correlates strongly with task performance, providing insights into their structure and potential training improvements.

Contribution

It reveals that embedding space structure, measured by ICA and neighbor overlap, predicts model performance across diverse tasks and languages.

Findings

Embedding space organization correlates with task performance (up to 0.97).

Different tasks vary in linearity and local information retention.

Embedding structure insights suggest new training objectives.

Abstract

In this paper, we show that high-performing embedding models organize their embedding spaces in a consistent way. We evaluate 25 contemporary embedding models on five MTEB tasks spanning four diverse task categories (retrieval, bitext mining, pair classification, and summarization) in both English and multilingual settings, and reveal that nearest-neighbor overlap and magnitude differences in independent component analysis (ICA) between paired text instances strongly correlate (even up to 0.97) with performance on the given task. Ultimately, we show that embedding tasks display varying degrees of linearity and reliance on retention of local information. Our results further the understanding of embeddings, their relation to model performance, and shed light on possible future training objectives and optimizing conditional embeddings.