arguana_neomme_260m_li

Multi-vector (late-interaction) embeddings of BEIR arguana (beir/arguana), encoded with Hcompany/NeoMME-260M-Retriever-ST-late at revision 023be2a8ab9d797f5aa76f5bf8b5dde78d819659.

Source data: ir_datasets beir/arguana (ir_datasets 0.6.3), which downloads arguana.zip (md5 8ad3e3c2a5867cdced806d6503f29b99). BEIR also publishes this corpus on the Hub as BeIR/arguana, 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)	[1,687,416, 128]	every document vector, concatenated document by document (412.0 MiB)
doclens.npy	int32	[8,674]	vectors per document; cumsum gives offsets
token_ids.npy	uint32	[1,687,416]	tokenizer id of each documents.npy row, 1:1
doc_ids.npy	<U47	[8,674]	original document ids
queries.npy	float32 (<f4)	[1,406, 1042, 128]	query vectors, zero-padded at the end (715.4 MiB)
query_lens.npy	int32	[1,406]	true vectors per query, before padding
queries_ids.npy	<U41	[1,406]	original query ids
qrels.test.tsv	text	1,406 rows	TREC qrels, qid \t 0 \t docid \t relevance, no header
gt_top100.tsv	text	140,600 rows	exact MaxSim top-100, see below

All positional indices (gt_top100.tsv, and the row order of every .npy file) refer to the order of doc_ids.npy and queries_ids.npy. Reordering either file invalidates gt_top100.tsv.

Statistics

documents	8,674
document vectors	1,687,416
vectors per document (min / median / mean / max)	2 / 173 / 194.5 / 1321
queries	1,406
vectors per query (min / median / max)	57 / 214 / 1042
queries with at least one qrel	1,406
qrels rows	1,406
embedding dimension	128

Encoding

model	Hcompany/NeoMME-260M-Retriever-ST-late
model revision	023be2a8ab9d797f5aa76f5bf8b5dde78d819659
library	sentence-transformers 6.0.1 MultiVectorEncoder (transformers 5.3.0, torch 2.11.0+cu130)
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	none (document_length unset; model limit 16,384 tokens, longest document here 1,321 vectors, so nothing hit it)
query truncation	none (checkpoint default; query_length unset)
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_top100.tsv

Exact brute-force MaxSim top-100 per query over the full corpus, from the vectors in this repo. 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.

Self-matches are included. 1,298 of 1,406 queries are themselves corpus documents with the same id and retrieve that document (typically at rank 1). This is the raw nearest-neighbour list; BEIR's evaluation drops such pairs (ignore_identical_ids), so exclude them before scoring against qrels.

Retrieval quality

Sanity check of the vectors, not a leaderboard number: exact MaxSim over the full corpus scored against qrels.test.tsv with ir_measures, with query-id == doc-id pairs dropped as BEIR does.

nDCG@10	Recall@100	MRR@10	MAP
0.4163	0.9566	0.3339	0.3465

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, 1042, 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 (1687416, 128)
• ✅ doclens.npy dtype/shape — <i4 (8674,)
• ✅ doc_ids.npy is a string array — <U47 (8674,)
• ✅ queries.npy dtype/shape — <f4 (1406, 1042, 128)
• ✅ query_lens.npy dtype/shape — <i4 (1406,)
• ✅ queries_ids.npy is a string array — <U41 (1406,)
• ✅ sum(doclens) == n_tokens — 1687416 vs 1687416
• ✅ no empty documents — min doclen 2
• ✅ len(doc_ids) == len(doclens) == corpus size — 8674, 8674, 8674
• ✅ doc_ids unique
• ✅ query arrays aligned — 1406, 1406, 1406
• ✅ doc and query dim agree — 128 / 128
• ✅ token_ids.npy dtype/shape — <u4 (1687416,)
• ✅ document vectors unit-norm (100k sample) — norm range [0.9995, 1.0005]
• ✅ query vectors unit-norm — norm range [1.000000, 1.000000]
• ✅ all vectors finite
• ✅ gt_top100.tsv has k rows per query — 140600 rows, k=100
• ✅ gt rows grouped by qidx with ranks 1..k and descending scores
• ✅ gt indices in range

Provenance

exported	2026-09-24
hardware	cpu