lance-format/ms-marco-v2.1-lance

MS MARCO v2.1 QA (Lance Format)

Lance-formatted version of MS MARCO v2.1 — Microsoft's machine reading comprehension benchmark — with MiniLM query embeddings stored inline alongside the candidate passages and human-written answers.

Why this version?

• One self-contained Lance dataset with ~900 k queries; each row is a query, the 10 candidate passages retrieved by Bing, the relevance flags, and the human-written reference answers.
• Pre-computed query embeddings (sentence-transformers/all-MiniLM-L6-v2, 384-dim, L2-normalized) with an IVF_PQ index — semantic query lookup without re-embedding.
• Full-text inverted indices on the query and the first selected passage.
• Designed for both retrieval research (use the index) and RAG / answer eval (use the passage list + answers).

Splits

Split	Rows
train.lance	808,731
validation.lance	101,093

Schema

Column	Type	Notes
query_id	int64	MS MARCO query id
query	string	The user's natural-language query
query_type	string	One of DESCRIPTION, NUMERIC, ENTITY, LOCATION, PERSON
answers	list<string>	Human-written reference answers
well_formed_answers	list<string>	Reference answers re-written as full sentences
passage_text	list<string>	Up to 10 candidate passages
passage_url	list<string>	Source URLs for each candidate
passage_is_selected	list<int8>	1 if Bing labelled the passage relevant
selected_passage	string?	First relevant passage (null if none)
query_emb	fixed_size_list<float32, 384>	MiniLM embedding of query (cosine-normalized)

Pre-built indices

• IVF_PQ on query_emb — metric=cosine
• INVERTED on query and selected_passage
• BTREE on query_id
• BITMAP on query_type

Quick start

import lance

ds = lance.dataset("hf://datasets/lance-format/ms-marco-v2.1-lance/data/validation.lance")
print(ds.count_rows(), ds.schema.names, ds.list_indices())

Semantic query lookup

import lance
import pyarrow as pa
from sentence_transformers import SentenceTransformer

encoder = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2", device="cuda")
q = encoder.encode(["how to compute determinant of a 3x3 matrix"], normalize_embeddings=True)[0]

ds = lance.dataset("hf://datasets/lance-format/ms-marco-v2.1-lance/data/validation.lance")
emb_field = ds.schema.field("query_emb")
hits = ds.scanner(
    nearest={"column": "query_emb", "q": pa.array([q.tolist()], type=emb_field.type)[0], "k": 5, "nprobes": 16, "refine_factor": 30},
    columns=["query_id", "query", "selected_passage", "answers"],
).to_table().to_pylist()
for h in hits:
    print(h["query"])
    print("  selected:", (h.get("selected_passage") or "")[:120])

Get all candidate passages for a query

import lance
ds = lance.dataset("hf://datasets/lance-format/ms-marco-v2.1-lance/data/validation.lance")
row = ds.scanner(filter="query_id = 1185869", columns=["query", "passage_text", "passage_is_selected"]).to_table().to_pylist()[0]
for text, sel in zip(row["passage_text"], row["passage_is_selected"]):
    print("[selected]" if sel else "[other]", text[:120])

Filter by query_type

ds = lance.dataset("hf://datasets/lance-format/ms-marco-v2.1-lance/data/train.lance")
numeric = ds.scanner(filter="query_type = 'NUMERIC'", columns=["query"], limit=5).to_table()

Why Lance?

• One dataset carries queries + passages + answers + embeddings + indices — no sidecar files.
• On-disk vector and full-text indices live next to the data, so search works on local copies and on the Hub.
• Schema evolution: add columns (alternate embeddings, generated answers, model predictions) without rewriting the data.

Source & license

Converted from microsoft/ms_marco (v2.1). MS MARCO is released under the MIT license.

Citation

@article{nguyen2016ms,
  title={MS MARCO: A Human Generated MAchine Reading COmprehension Dataset},
  author={Nguyen, Tri and Rosenberg, Mir and Song, Xia and Gao, Jianfeng and Tiwary, Saurabh and Majumder, Rangan and Deng, Li},
  journal={arXiv preprint arXiv:1611.09268},
  year={2016}
}