Do We Really Need Specialization? Evaluating Generalist Text Embeddings for Zero-Shot Recommendation and Search

TL;DR

This paper demonstrates that large-scale pre-trained generalist text embedding models can effectively perform zero-shot recommendation and search tasks without domain-specific fine-tuning, outperforming traditional and fine-tuned models.

Contribution

The study shows that generalist text embeddings pre-trained on large corpora can achieve strong zero-shot performance in recommendation and search, challenging the need for task-specific fine-tuning.

Findings

GTEs outperform fine-tuned models in zero-shot recommendation and search

Dimensionality reduction via PCA improves model performance

GTEs distribute features more evenly across embedding space

Abstract

Pre-trained language models (PLMs) are widely used to derive semantic representations from item metadata in recommendation and search. In sequential recommendation, PLMs enhance ID-based embeddings through textual metadata, while in product search, they align item characteristics with user intent. Recent studies suggest task and domain-specific fine-tuning are needed to improve representational power. This paper challenges this assumption, showing that Generalist Text Embedding Models (GTEs), pre-trained on large-scale corpora, can guarantee strong zero-shot performance without specialized adaptation. Our experiments demonstrate that GTEs outperform traditional and fine-tuned models in both sequential recommendation and product search. We attribute this to a superior representational power, as they distribute features more evenly across the embedding space. Finally, we show that compressing embedding dimensions by focusing on the most informative directions (e.g., via PCA) effectively reduces noise and improves the performance of specialized models. To ensure reproducibility, we provide our repository at https://split.to/gte4ps.