scifact_mlateon

Multi-vector (late-interaction) embeddings of BEIR scifact (beir/scifact/test), encoded with lightonai/mLateOn at revision edd378f99593c0ac8a15518b97ad89786b02685e.

Source data: ir_datasets beir/scifact/test (ir_datasets 0.6.3), which downloads scifact.zip (md5 5f7d1de60b170fc8027bb7898e2efca1). BEIR also publishes this corpus on the Hub as BeIR/scifact, whose card gives this dataset's license; the data here was loaded through ir_datasets, not from that repo. Document, query and qrel ids are the source's own ids, unchanged.

Every document is one variable-length set of 128-d vectors; every query is one variable-length set of 128-d vectors. Documents and queries are stored at different precisions (fp16 and fp32 respectively), see Encoding.

Files

file	dtype	shape	contents
documents.npy	float16 (<f2)	[1,629,339, 128]	every document vector, concatenated document by document (397.8 MiB)
doclens.npy	int32	[5,183]	vectors per document; cumsum gives offsets
token_ids.npy	uint32	[1,629,339]	tokenizer id of each documents.npy row, 1:1
doc_ids.npy	<U9	[5,183]	original document ids
queries.npy	float32 (<f4)	[300, 53, 128]	query vectors, zero-padded at the end (7.8 MiB)
query_lens.npy	int32	[300]	true vectors per query, before padding
queries_ids.npy	<U4	[300]	original query ids
qrels.test.tsv	text	339 rows	TREC qrels, qid \t 0 \t docid \t relevance, no header
gt_top1000.tsv	text	300,000 rows	exact MaxSim top-1000, see below
gt_top100.tsv	text	30,000 rows	first 100 ranks of gt_top1000.tsv, same format

All positional indices (the gt_top*.tsv files, and the row order of every .npy file) refer to the order of doc_ids.npy and queries_ids.npy. Reordering either file invalidates the ground truth.

Statistics

documents	5,183
document vectors	1,629,339
vectors per document (min / median / mean / max)	69 / 288 / 314.4 / 1902
queries	300
vectors per query (min / median / mean / max)	8 / 19 / 21.0 / 53
queries with at least one qrel	300
qrels rows	339
embedding dimension	128

Encoding

model	lightonai/mLateOn
model revision	edd378f99593c0ac8a15518b97ad89786b02685e
library	sentence-transformers 6.1.0 MultiVectorEncoder (transformers 5.17.0, torch 2.13.0+cu126)
document compute dtype	float16 (model weights loaded at this dtype for the document pass)
document storage dtype	fp16
query compute dtype	float32 (model weights loaded at this dtype for the query pass)
query storage dtype	fp32
normalization	L2, by the model's own Normalize module, before the storage cast
document truncation	8,192 tokens (the checkpoint's document_length). No document reached it; longest here 1,902 vectors
query truncation	8,192 tokens (the checkpoint's query_length)
document skiplist	none (empty skiplist_words): every document token is kept, so doclens is the real token count
document input	title + "\n\n" + text when the corpus has a title, else text; stripped
query input	query text, stripped of surrounding whitespace, formatted by the model's own query prompt/template
query vectors	every vector the model emits for the query is kept, including any query-expansion tokens its template adds; query_lens counts them all
document padding	none: documents.npy holds real vectors only, sum(doclens) == n_tokens
query padding	rows at or beyond query_lens[i] in queries.npy[i] are exactly zero
token_ids	tokenizer id of each kept document token (no skiplist, so every token), aligned 1:1 with documents.npy

Ground truth: gt_top1000.tsv and gt_top100.tsv

Exact brute-force MaxSim top-1000 per query over the full corpus, from the vectors in this repo. gt_top100.tsv holds the first 100 ranks per query of the same lists (the original layout of these exports).

No header; tab-separated qidx docidx rank score:

• qidx: 0-based row into queries_ids.npy / queries.npy
• docidx: 0-based position into doc_ids.npy / doclens.npy
• rank: 1-based, descending score
• score: sum over the query's query_lens[qidx] vectors of max over the document's vectors of the dot product, computed in fp32 with the fp16 document vectors upcast to fp32. Expansion vectors are included in the sum. Printed to 6 decimals.

Retrieval quality

Sanity check of the vectors, not a leaderboard number: gt_top1000.tsv (exact MaxSim over the full corpus) scored against qrels.test.tsv with ir_measures.

nDCG@10	MRR@10	Success@5	Recall@100	Recall@1000	MAP@1000
0.7605	0.7305	0.8367	0.9733	1.0000	0.7270

Loading

import numpy as np

documents = np.load("documents.npy", mmap_mode="r")      # [n_tokens, 128] float16
doclens = np.load("doclens.npy")                         # [n_docs] int32
offsets = np.concatenate([[0], np.cumsum(doclens)])
doc_ids = np.load("doc_ids.npy")                         # [n_docs] str

def document(i):
    return documents[offsets[i]:offsets[i + 1]]           # [doclens[i], 128]

queries = np.load("queries.npy")                         # [n_queries, 53, 128] float32
query_lens = np.load("query_lens.npy")                   # [n_queries] int32
query_ids = np.load("queries_ids.npy")                   # [n_queries] str

def query(j):
    return queries[j, :query_lens[j]]                     # [query_lens[j], 128]

def maxsim(q, d):
    return (q @ d.astype(np.float32).T).max(axis=1).sum()

Validation

Checks run by the exporter on the files exactly as written here:

• ✅ file set — missing=[] extra=[]
• ✅ documents.npy dtype/shape — <f2 (1629339, 128)
• ✅ doclens.npy dtype/shape — <i4 (5183,)
• ✅ doc_ids.npy is a string array — <U9 (5183,)
• ✅ queries.npy dtype/shape — <f4 (300, 53, 128)
• ✅ query_lens.npy dtype/shape — <i4 (300,)
• ✅ queries_ids.npy is a string array — <U4 (300,)
• ✅ sum(doclens) == n_tokens — 1629339 vs 1629339
• ✅ no empty documents — min doclen 69
• ✅ len(doc_ids) == len(doclens) == corpus size — 5183, 5183, 5183
• ✅ doc_ids unique
• ✅ query arrays aligned — 300, 300, 300
• ✅ doc and query dim agree — 128 / 128
• ✅ token_ids.npy dtype/shape — <u4 (1629339,)
• ✅ document vectors unit-norm (100k sample) — norm range [0.9994, 1.0006]
• ✅ query vectors unit-norm — norm range [1.000000, 1.000000]
• ✅ all vectors finite
• ✅ gt_top100.tsv has k rows per query — 30000 rows, k=100
• ✅ gt_top100.tsv rows grouped by qidx with ranks 1..k and descending scores
• ✅ gt_top100.tsv indices in range
• ✅ gt_top1000.tsv has k rows per query — 300000 rows, k=1000
• ✅ gt_top1000.tsv rows grouped by qidx with ranks 1..k and descending scores
• ✅ gt_top1000.tsv indices in range
• ✅ gt_top100.tsv is the first 100 ranks of gt_top1000.tsv

Provenance

exported	2026-09-25
hardware	Tesla V100S-PCIE-32GB
revised	2026-09-29: ground truth extended to top-1000 (gt_top1000.tsv, exact MaxSim over this repo's vectors on Tesla V100S-PCIE-32GB); gt_top100.tsv rewritten as its first 100 ranks: 28 rows differ from the previous file, all of them documents with identical scores listed in a different order