REFINE on Scarce Data: Retrieval Enhancement through Fine-Tuning via Model Fusion of Embedding Models

TL;DR

This work proposes REFINE, a novel technique that generates synthetic data from available documents and then uses a model fusion approach to fine-tune embeddings for improved retrieval performance in new domains, while preserving out-of-domain capability.

Abstract

Retrieval augmented generation (RAG) pipelines are commonly used in tasks such as question-answering (QA), relying on retrieving relevant documents from a vector store computed using a pretrained embedding model. However, if the retrieved context is inaccurate, the answers generated using the large language model (LLM) may contain errors or hallucinations. Although pretrained embedding models have advanced, adapting them to new domains remains challenging. Fine-tuning is a potential solution, but industry settings often lack the necessary fine-tuning data. To address these challenges, we propose REFINE, a novel technique that generates synthetic data from available documents and then uses a model fusion approach to fine-tune embeddings for improved retrieval performance in new domains, while preserving out-of-domain capability. We conducted experiments on the two public datasets: SQUAD and RAG-12000 and a proprietary TOURISM dataset. Results demonstrate that even the standard fine-tuning with the proposed data augmentation technique outperforms the vanilla pretrained model. Furthermore, when combined with model fusion, the proposed approach achieves superior performance, with a 5.76% improvement in recall on the TOURISM dataset, and 6.58 % and 0.32% enhancement on SQUAD and RAG-12000 respectively.

Figures (2)

• Figure 1: (A) standard BGE process followed for fine-tuning. (B) training process for REFINE.
• Figure 2: Recall at different top_k values for vanilla (pretrained) BGE, fine-tuned (FT) BGE, and REFINE.