Following the Autoregressive Nature of LLM Embeddings via Compression and Alignment

TL;DR

This paper introduces AutoRegEmbed, a contrastive learning method that aligns LLM embeddings with their conditional probability distributions, improving semantic encoding and outperforming traditional contrastive approaches.

Contribution

The paper presents a novel contrastive learning framework that integrates information compression and distribution alignment to better utilize LLM embeddings.

Findings

AutoRegEmbed outperforms traditional contrastive methods.

Achieves performance comparable to state-of-the-art models.

Effectively captures global semantics in embeddings.

Abstract

A new trend uses LLMs as dense text encoders via contrastive learning. However, since LLM embeddings predict the probability distribution of the next token, they are inherently generative and distributive, conflicting with contrastive learning, which requires embeddings to capture full-text semantics and align via cosine similarity. This discrepancy hinders the full utilization of LLMs' pre-training capabilities, resulting in inefficient learning. In response to this issue, we propose AutoRegEmbed, a new contrastive learning method built on embedding conditional probability distributions, which integrates two core tasks: information compression and conditional distribution alignment. The information compression task encodes text into the embedding space, ensuring that the embedding vectors capture global semantics. The conditional distribution alignment task focuses on aligning text embeddings with positive samples embeddings by leveraging the conditional distribution of embeddings while simultaneously reducing the likelihood of generating negative samples from text embeddings, thereby achieving embedding alignment and uniformity. Experimental results demonstrate that our method significantly outperforms traditional contrastive learning approaches and achieves performance comparable to state-of-the-art models when using the same amount of data.