English

Beyond instruction-conditioning, MoTE: Mixture of Task Experts for Multi-task Embedding Models

Machine Learning 2025-06-24 v1 Computation and Language

Abstract

Dense embeddings are fundamental to modern machine learning systems, powering Retrieval-Augmented Generation (RAG), information retrieval, and representation learning. While instruction-conditioning has become the dominant approach for embedding specialization, its direct application to low-capacity models imposes fundamental representational constraints that limit the performance gains derived from specialization. In this paper, we analyze these limitations and introduce the Mixture of Task Experts (MoTE) transformer block, which leverages task-specialized parameters trained with Task-Aware Contrastive Learning (\tacl) to enhance the model ability to generate specialized embeddings. Empirical results show that MoTE achieves 64%64\% higher performance gains in retrieval datasets (+3.27+5.21+3.27 \rightarrow +5.21) and 43%43\% higher performance gains across all datasets (+1.81+2.60+1.81 \rightarrow +2.60). Critically, these gains are achieved without altering instructions, training data, inference time, or number of active parameters.

Keywords

Cite

@article{arxiv.2506.17781,
  title  = {Beyond instruction-conditioning, MoTE: Mixture of Task Experts for Multi-task Embedding Models},
  author = {Miguel Romero and Shuoyang Ding and Corey D. Barret and Georgiana Dinu and George Karypis},
  journal= {arXiv preprint arXiv:2506.17781},
  year   = {2025}
}
R2 v1 2026-07-01T03:27:57.985Z