F2LLM-v2: Inclusive, Performant, and Efficient Embeddings for a Multilingual World

TL;DR

F2LLM-v2 introduces a family of multilingual embedding models that are highly efficient, support over 200 languages, and achieve top performance on multiple benchmarks, with open-source availability.

Contribution

The paper presents a new scalable, efficient, and multilingual embedding model family with innovative training techniques and extensive evaluation, advancing open-source multilingual NLP resources.

Findings

F2LLM-v2-14B ranks first on 11 MTEB benchmarks.

Smaller models achieve state-of-the-art results for resource-constrained tasks.

Models support over 200 languages, including underserved ones.

Abstract

We present F2LLM-v2, a new family of general-purpose, multilingual embedding models in 8 distinct sizes ranging from 80M to 14B. Trained on a newly curated composite of 60 million publicly available high-quality data samples, F2LLM-v2 supports more than 200 languages, with a particular emphasis on previously underserved mid- and low-resource languages. By integrating a two-stage LLM-based embedding training pipeline with matryoshka learning, model pruning, and knowledge distillation techniques, we present models that are far more efficient than previous LLM-based embedding models while retaining competitive performances. Extensive evaluations confirm that F2LLM-v2-14B ranks first on 11 MTEB benchmarks, while the smaller models in the family also set a new state of the art for resource-constrained applications. To facilitate open-source embedding model research, we release all models, data, code, and intermediate checkpoints.