Korean Sparse & Multi-Vector Retrievers

SOTA Korean sparse and multi-vector models for dense-sparse hybrid retrieval.

Trained and open-sourced Korean Sparse and Multi-Vector models, achieving State-of-the-Art performance among corresponding architectures (as of Feb. 2026) on the Korean Retrieval Benchmark.

  • Outperformed existing multilingual and Korean fine-tuned models
  • Provided highly optimized, reproducible pipelines to advance dense-sparse hybrid retrieval experiments
colbert-ko-v1 splade-ko-v1 inference-free-splade-ko-v1