Refining Sentence Embedding Model through Ranking Sentences Generation with Large Language Models

TL;DR

This paper introduces a novel method that uses controlled large language model generation to enhance sentence embeddings by incorporating ranking information, achieving state-of-the-art results with efficient synthesis.

Contribution

It proposes a controlled generation technique in LLMs to incorporate ranking information into sentence embeddings, improving semantic distinctions and performance.

Findings

Achieves new state-of-the-art on multiple benchmarks.

Demonstrates effective integration of ranking information.

Maintains modest computational cost.

Abstract

Sentence embedding is essential for many NLP tasks, with contrastive learning methods achieving strong performance using annotated datasets like NLI. Yet, the reliance on manual labels limits scalability. Recent studies leverage large language models (LLMs) to generate sentence pairs, reducing annotation dependency. However, they overlook ranking information crucial for fine-grained semantic distinctions. To tackle this challenge, we propose a method for controlling the generation direction of LLMs in the latent space. Unlike unconstrained generation, the controlled approach ensures meaningful semantic divergence. Then, we refine exist sentence embedding model by integrating ranking information and semantic information. Experiments on multiple benchmarks demonstrate that our method achieves new SOTA performance with a modest cost in ranking sentence synthesis.