colbert-ko-en-v2

colbert-ko-en-v2 is a Korean-English bilingual late-interaction (ColBERT) retriever built on skt/A.X-Encoder-base. It encodes every token into a 128-dimensional vector and scores query–document pairs with MaxSim, keeping the term-level matching signal that a single pooled vector discards.

At 149M parameters it reaches an average nDCG@10 of 0.7984 across the nine MTEB(kor, v2) retrieval tasks — the strongest late-interaction model on that suite, and in the same range as single-vector models many times its size.

Key Characteristics

• Late interaction: one 128-d vector per token, scored with MaxSim. No pooling.
• Long documents: up to 8,192 tokens, on ModernBERT's RoPE attention.
• Compact: 149M parameters.
• No instruction prefixes: queries and documents need no task instruction. Query expansion to 64 tokens is handled by the model.

Usage

pip install -U pylate

Indexing documents

from pylate import indexes, models, retrieve

model = models.ColBERT(model_name_or_path="yjoonjang/colbert-ko-en-v2")

index = indexes.PLAID(
    index_folder="pylate-index",
    index_name="index",
    override=True,
)

documents_ids = ["1", "2", "3"]
documents = [
    "세종대왕은 1443년에 훈민정음을 창제하고 1446년에 이를 반포하였다.",
    "김치는 배추나 무를 소금에 절인 뒤 고춧가루와 젓갈을 넣어 발효시킨 음식이다.",
    "한라산은 해발 1,947m로 남한에서 가장 높은 산이며 제주도 중앙에 자리한다.",
]

documents_embeddings = model.encode(
    documents,
    batch_size=32,
    is_query=False,
    show_progress_bar=True,
)

index.add_documents(
    documents_ids=documents_ids,
    documents_embeddings=documents_embeddings,
)

To reuse an existing index, instantiate it without override:

index = indexes.PLAID(index_folder="pylate-index", index_name="index")

Retrieving top-k documents

retriever = retrieve.ColBERT(index=index)

queries_embeddings = model.encode(
    ["훈민정음은 언제 만들어졌나요?"],
    batch_size=32,
    is_query=True,
    show_progress_bar=True,
)

scores = retriever.retrieve(
    queries_embeddings=queries_embeddings,
    k=10,
)

Reranking

To rerank a first-stage candidate list without building an index:

from pylate import models, rank

model = models.ColBERT(model_name_or_path="yjoonjang/colbert-ko-en-v2")

queries = [
    "전기차 폐배터리는 어떻게 재활용하나요?",
    "겨울에 한라산을 오를 때 필요한 장비는?",
]
documents = [
    [
        "폐배터리에서 리튬과 코발트를 회수하는 습식 제련 공정이 상용화되고 있다.",
        "급속 충전기는 30분 내외로 배터리를 80%까지 충전할 수 있다.",
    ],
    [
        "겨울 한라산 산행에는 아이젠과 방한 장갑이 필수이며 입산 시간이 제한된다.",
        "제주 올레길은 해안을 따라 이어지는 27개 코스로 구성되어 있다.",
        "적설기에는 등산화에 스패츠를 착용해 눈이 들어가는 것을 막는 것이 좋다.",
    ],
]
documents_ids = [[1, 2], [1, 3, 2]]

queries_embeddings = model.encode(queries, is_query=True)
documents_embeddings = model.encode(documents, is_query=False)

reranked_documents = rank.rerank(
    documents_ids=documents_ids,
    queries_embeddings=queries_embeddings,
    documents_embeddings=documents_embeddings,
)

Evaluation

nDCG@10 on the nine MTEB(kor, v2) retrieval tasks.

Late-interaction rows were measured with mteb 2.18.16 and PLAID retrieval; single-vector rows are taken from the official MTEB results repository, except for Belebele using only Korean query, Korean corpus subset. The original version also includes cross-lingual subsets (Korean query - English corpus, English query - Korean corpus).

Model Details

• Model type: ColBERT / PyLate
• Base model: skt/A.X-Encoder-base — ModernBERT, 22 layers, 768 hidden
• Parameters: 149M
• Output dimensionality: 128 per token
• Similarity function: MaxSim
• Document length: 8,192 tokens
• Query length: 64 tokens (MASK expansion)
• Languages: Korean, English

ColBERT(
  (0): Transformer({'max_seq_length': 8191, 'do_lower_case': False, 'architecture': 'ModernBertModel'})
  (1): Dense({'in_features': 768, 'out_features': 128, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity', 'use_residual': False})
)

Citation

@misc{colbert-ko-en-v2,
  title  = {colbert-ko-en-v2: a Korean-English bilingual late-interaction retriever},
  author = {Jang, Yongjoon},
  year   = {2026},
  url    = {https://huggingface.co/yjoonjang/colbert-ko-en-v2},
}

@inproceedings{santhanam-etal-2022-colbertv2,
  title     = {ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction},
  author    = {Santhanam, Keshav and Khattab, Omar and Saad-Falcon, Jon and Potts, Christopher and Zaharia, Matei},
  booktitle = {Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies},
  year      = {2022},
  pages     = {3715--3734},
}

@misc{PyLate,
  title  = {PyLate: Flexible Training and Retrieval for Late Interaction Models},
  author = {Chaffin, Antoine and Sourty, Raphaël},
  year   = {2024},
  url    = {https://github.com/lightonai/pylate},
}