Granite Embedding Models
This work addresses retrieval and search challenges for enterprise users, offering high-quality, publicly available models, but it is incremental as it builds on existing techniques like contrastive finetuning and knowledge distillation.
The paper tackles the problem of improving retrieval tasks by introducing Granite Embedding models, a family of encoder-based models for dense and sparse retrieval in English and multilingual settings, and reports that they significantly outperform publicly available models of similar sizes on internal IBM tasks while achieving equivalent performance on standard benchmarks.
We introduce the Granite Embedding models, a family of encoder-based embedding models designed for retrieval tasks, spanning dense-retrieval and sparse retrieval architectures, with both English and Multilingual capabilities. This report provides the technical details of training these highly effective 12 layer embedding models, along with their efficient 6 layer distilled counterparts. Extensive evaluations show that the models, developed with techniques like retrieval oriented pretraining, contrastive finetuning, knowledge distillation, and model merging significantly outperform publicly available models of similar sizes on both internal IBM retrieval and search tasks, and have equivalent performance on widely used information retrieval benchmarks, while being trained on high-quality data suitable for enterprise use. We publicly release all our Granite Embedding models under the Apache 2.0 license, allowing both research and commercial use at https://huggingface.co/collections/ibm-granite.