Sentence Transformers v6.0 Introduces Multi-Vector Late Interaction Retrieval

Sentence Transformers v6.0 Introduces Multi-Vector Late Interaction Retrieval Sentence Transformers v6.0 adds native support for ColBERT-style late interaction retrieval via MultiVectorEncoder, enabling token-level MaxSim scoring across text, visual documents, audio, and video. MultiVectorEncoder replaces single-vector text compression by retaining low-dimensional contextualized token embeddings scored via the MaxSim operator. Late interaction achieves higher retrieval accuracy on long-passage and evidence-specific queries, though it substantially increases vector storage requirements. Storage and computational overheads can be managed using PLAID compression, hierarchical token pooling, or bi-encoder retrieve-and-rerank pipelines. Native multi-vector indexing is supported across databases such as Qdrant, Weaviate, Vespa, and LanceDB, with support extending to multimodal models like ColPali and ColQwen. Mechanics of Multi-Vector Late Interaction Sentence Transformers v6.0 introduces MultiVectorEncoder to implement ColBERT-style late interaction retrieval, supporting checkpoints from PyLate, Stanford-NLP ColBERT, and colpali-engine. Rather than condensing text into a single embedding vector, multi-vector models project contextualized token embeddings into lower dimensions (typically 128) and keep all tokens. In late interaction, documents are encoded independently offline, while query-time retrieval scores texts using the MaxSim operator, which calculates the sum of maximum dot products between each query token and any document token. Because token embeddings are L2-normalized, dot products reflect cosine similarity, enabling soft semantic alignment across distinct terms as well as exact keyword matching. This approach improves retrieval quality on long passages and complex multi-requirement queries, where models like mLateOn scored 77.92 on the MLDR benchmark compared to 51.59 for mDenseOn. Storage Demands, Compression, and Reranking Retaining per-token vectors introduces significant storage overhead compared to standard dense bi-encoders. On 4,874 Natural Questions passages, lightonai/LateOn produced 608,414 token vectors requiring 311.5 MB in float32, whereas all-MiniLM-L6-v2 occupied only 7.5 MB. To mitigate index bloat, fast-plaid compression uses centroids and quantized residuals to compress the same corpus to 92 MB while maintaining top-rank ordering. Additionally, HierarchicalTokenPooling clusters document token vectors using Ward linkage on cosine distance, cutting vector counts by 2x to 4x while retaining 99% to 100.6% of unpooled BEIR benchmark performance. Multi-vector models can also operate in a retrieve-and-rerank setup, where a standard dense bi-encoder handles initial candidate retrieval before the multi-vector model rescores top candidates via batched matrix multiplication, avoiding late-interaction index maintenance entirely. Vector Database Ecosystem and Performance Several vector databases provide native indexing and scoring for multi-vectors, including Qdrant (v1.10+), Weaviate (v1.29+), Vespa, LanceDB (v0.15.0+), and Milvus (v2.6.4+). Benchmarks on 4,874 passages showed fast-plaid indexing in 5 seconds on GPU with 11 ms query latencies. Qdrant ingested the corpus in 26.3 seconds on CPU and executed queries in 18 ms with exact MaxSim scoring. Weaviate ingested the dataset in 41 seconds on CPU with 17 ms queries, though its accelerated MUVERA encoding led to accuracy degradation and missed relevant top passages. Vespa deploys custom tensor expressions for exact MaxSim calculation and offers a phased ranking app that stores int8-binarized vectors for candidate pruning prior to second-phase MaxSim rescoring. Multimodal Retrieval and Framework Features MultiVectorEncoder outputs consistent token-vector matrices regardless of input type, enabling visual and cross-modal retrieval across compatible databases. Vision-language models like colqwen2.5-v0.2 match text query tokens directly against image patch vectors without an OCR step, supporting interpretability heatmaps that map score contributions back to visual regions. Larger models like colqwen-omni-v0.1 extend zero-shot retrieval to raw 16 kHz mono audio waveforms and sampled video frames without transcription. Sentence Transformers v6.0 also incorporates inference backends including fp16 with Flash Attention, OpenVINO, and int8 quantization, and provides automated evaluation via MultiVectorNanoBEIREvaluator.

Sentence Transformers 6.0支持ColBERT多向量迟交互检索 Sentence Transformers v6.0推出了MultiVectorEncoder，全面支持ColBERT风格的多向量迟交互检索机制。该架构保留每个Token的独立向量并结合MaxSim算子进行细粒度匹配，在显著提升长文档与特定证据检索精度的同时，配合分层Token池化与数据库索引方案解决存储膨胀难题。 Sentence Transformers v6.0引入MultiVectorEncoder，通过保留Token级独立向量与MaxSim算子实现细粒度迟交互匹配。 针对多向量存储开销过大的问题，支持fast-plaid残差量化和分层Token池化算法，可在几乎不损精度的前提下减少2至5倍向量体积。 支持Qdrant、Weaviate、Vespa等主流向量数据库原生索引，并可作为比交叉编码器更高效的两阶段重排序器。 兼容ColPali等视觉与全模态模型，支持无需OCR的图表、原始音频与视频切片检索，且评分可直接分解映射为可解释性热力图。 迟交互机制与MaxSim打分 传统密集模型将整个文本压缩为单一向量，而多向量模型跳过全篇压缩，将每个Token的上下文表征映射至较低维度（通常为128维）。在检索阶段，文档在离线状态下独立编码，查询则与文档计算MaxSim相似度：即针对每一个查询Token，寻找与其点积最大的文档Token，并将所有最大值求和。由于Token向量经过L2归一化，点积即为余弦相似度，使模型能够捕捉非完全重合词汇之间的软对齐。在长文档MLDR基准测试中，多语言多向量模型mLateOn取得了77.92分，大幅高于密集模型mDenseOn的51.59分。 存储瓶颈与Token压缩优化 多向量检索的主要代价是极高的存储需求。对Natural Questions语料库的4,874个段落进行编码时，生成了608,414个Token向量，float32格式下占用311.5 MB，远高于传统单向量模型的7.5 MB至15 MB。为此，框架集成了基于聚类质心与量化残差的fast-plaid技术，将体积压缩至92 MB并维持顶层排序稳定。此外，基于Ward分层聚类的HierarchicalTokenPooling算法可在文档侧按设定的池化因子对相邻Token进行合并；实验表明，在池化因子为2时（体积缩减约2倍），模型在BEIR上的平均检索性能保留率可达100.6%。 向量数据库集成与两阶段重排 Qdrant、Weaviate、Vespa、LanceDB与VectorChord等向量引擎已原生支持多向量与MaxSim打分。测试显示，Qdrant在关闭图构建的情况下精确复现了MaxSim分值并实现18毫秒查询；Weaviate则通过MUVERA编码实现更快写入与查询，但存在一定的召回率下降。对于无需维护全量多向量索引的场景，多向量编码器适合在“先检索后重排”管道中充当第二阶段重排序器：先由普通双编码器筛选出前数十个候选，再由多向量编码器批量编码并执行矩阵乘法打分，其计算成本显著低于逐对前向计算的交叉编码器。 多模态扩展与分数可解释性 多向量编码器对文本、图像切片、音频和视频统一输出Token向量矩阵。以ColPali架构为例，模型可直接将查询文本Token与文档图像切片匹配，无需OCR预处理；全模态模型colqwen-omni-v0.1还能在无转录文本的情况下直接基于16 kHz单声道音频波形进行跨模态零样本检索。由于MaxSim总分可精确分解为各个查询Token与文档Token之间的对应最大贡献，框架支持生成基于图像切片或文本Token的归因热力图，提供高精度的可解释性。 基准评估与运行环境要求 Sentence Transformers v6.0原生兼容PyLate、Stanford-NLP与colpali-engine的模型权重，运行依赖transformers v5.x、torch 2.2+及huggingface-hub v1.x。在13个NanoBEIR子集评测中，128维的LateOn平均NDCG达到0.6868，优于768维DenseOn的0.6764，并在9个数据集上胜出。运行时支持Flash Attention加速，半精度下可获得2.44倍的吞吐提升；但对于包含非注意力查询扩展标记的特定ColBERT架构，加载时会自动降级为SDPA以保证扩展有效性。

Sentence Transformers is a Python library for using and training embedding and reranker models for applications like retrieval augmented generation, semantic search, and more. With the v6.0 update, it gains a fourth model type: MultiVectorEncoder , for ColBERT-style late interaction retrieval. Any PyLate checkpoint and any Stanford-NLP ColBERT checkpoint loads straight into it, and colpali-engine models for visual document retrieval can be used too, through the same familiar API you already use for dense, sparse, and reranker models. Where a regular embedding model compresses a whole text into one vector, a multi-vector model keeps one vector per token and scores query against document with the MaxSim operator. That preserves token-level matching information that a single vector has to average away, which usually means stronger retrieval at the cost of a bigger index. It's also the state of the art for visual document retrieval, where a text query is matched against page images directly, with no OCR step in between. In this blogpost, we'll show you how to use these models: loading the various checkpoint formats, encoding and scoring, plugging them into a search stack, running them on page images, and keeping the index affordable. Everything below runs on a plain pip install -U sentence-transformers . This blog is about using multi-vector models. If you want to learn how to train them, see the companion Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers blogpost. Table of Contents What are Multi-Vector Models? The MaxSim Operator What You Gain, and What It Costs Installation Loading a Model Inspecting What a Checkpoint Configured Encoding Queries and Documents Scoring with MaxSim Score Magnitude and MeanMaxSim Semantic Search Retrieve and Rerank Indexing Visual Document Retrieval Audio Retrieval Video Retrieval Interpretability Token Pooling Speeding Up Inference Evaluating a Model Coming from PyLate or colpali-engine Supported Models Acknowledgements Additional Resources What are Multi-Vector Models? A dense embedding model reads a text and returns a single fixed-size vector. Everything the model noticed has to fit in those 384, 768, or 1024 numbers, and similarity is one dot product between two such summaries. This works remarkably well, but the compression is lossy in a specific way: a rare entity, an exact identifier, or one crucial clause in a long passage all have to compete for room in the same vector. A query with several requirements at once runs into the same wall. For "green sofa with wooden legs and rounded cushions", a single vector has to blend all four into one point, so a green sofa with the wrong legs ends up sitting close to the one you actually asked for. A multi-vector model (also called a late-interaction or ColBERT-style model, after the ColBERT paper ) skips that compression. It runs the same transformer, but instead of pooling the token embeddings into one vector, it projects each token embedding down to a small dimension (classically 128) and keeps all of them. A 9-token document becomes a 9x128 matrix, not a 1x128 vector. The interaction between query and document is then deferred until scoring time, which is where the name "late interaction" comes from. A cross-encoder interacts early, with both texts going through the model together, which is accurate but leaves nothing to precompute, since every document has to be re-encoded for each new query. A bi-encoder, which is what the dense embedding model above is, barely interacts at all (one dot product between two finished summaries), and that is exactly what lets you encode a collection once and query it fast. Late interaction sits in between. Documents are still encoded independently and can be indexed offline, but scoring compares every query token against every document token, which leaves far more room for the two to interact. The MaxSim Operator Scoring uses MaxSim: for each query token, take its highest similarity against any document token, then sum those maxima across the query. MaxSim ( Q , D ) = ∑ Q i ∈ Q max ⁡ D j ∈ D Q i ⋅ D j \text{MaxSim}(Q, D) = \sum_{Q_i \in Q} \max_{D_j \in D} Q_i \cdot D_j Because the token embeddings are L2-normalized, each of those dot products is a cosine similarity in [-1, 1] , so the whole sum lands within [-num_query_tokens, num_query_tokens] . You can read the operator as a soft alignment, where every query token points at the one document token that best explains it, and the score is how well the document supports the query overall. The alignment doesn't have to be lexical, since the token embeddings are contextualized. Encode "Where do penguins live?" against "Penguins inhabit Antarctica." with lightonai/mLateOn and the query token live finds its best match on inhabit at 0.94, a word it shares no characters with! That is the thing lexical retrieval cannot do, BM25 and its relatives need the term itself, so synonyms and paraphrases slip past them. Dense embedding models bridge that gap as well, of course. What late interaction adds is that it does so without giving up the other direction. When an exact match is what matters (a product code, a surname, a function name), MaxSim still has that token sitting there on its own, where a single-vector model had to average it in with everything else. It isn't one-to-one either, since several query tokens routinely settle on the same document token. What You Gain, and What It Costs You gain retrieval quality, particularly on queries where one specific piece of a document is what makes it relevant, on multi-requirement queries like the sofa above where each requirement gets to find its own evidence, and on out-of-domain data where a dense model's compression was tuned for a different distribution. That compression is learned from the training queries, so the model learns to keep what they needed and drop everything else, which may include exactly what your production queries ask about. The effect grows with document length, since more text has to fit in the same fixed vector. The cost is index size. One vector per token instead of one vector per document is a lot more vectors, only partly offset by the smaller dimension. Encoding 4,874 Natural Questions passages with lightonai/LateOn produced 608,414 token vectors, an average of 124.8 per passage: Representation Vectors Dimensions float32 size Dense, all-MiniLM-L6-v2 4,874 384 7.5 MB Dense, gte-modernbert-base 4,874 768 15.0 MB Multi-vector, LateOn 608,414 128 311.5 MB That's about 42x the storage of the MiniLM index, or 62 KiB per passage. However, indexes are often compressed, e.g. the same 608,414 vectors take 92 MB as a fast-plaid index, since PLAID stores a centroid id plus a quantized residual per vector rather than the vector itself. For scale, a 4096-dimensional dense model like Qwen3-Embedding-8B would need about 80 MB for these same 4,874 passages, so a compressed multi-vector index sits in the same territory as the dense indexes people already run. Token Pooling cuts the vector count before any of that, and Retrieve and Rerank avoids building an index at all. PyLate comes up throughout this post, so here is the short version. Sentence Transformers handled dense and sparse models but not late interaction, so LightOn built PyLate on top of it to close that gap, adding the training, inference, and retrieval pieces these models need. Much of what you'll load below was trained with it, and LightOn built an ecosystem around it too, including fast-plaid , the late-interaction index that turns up in Indexing . With v6.0 those capabilities live in Sentence Transformers itself. With the tradeoff in mind, let's get a model running. Installation Multi-vector models work with a plain install: pip install -U sentence-transformers For ColPali-style visual document retrieval, you also need the image dependencies (see Installation for all extras, and Multimodal Embedding & Reranker Models for multimodal support in general): pip install -U "sentence-transformers[image]" Sentence Transformers v6.0 requires transformers v5.x, torch 2.2+, and huggingface-hub v1.x. If you pin any of those lower, plan the upgrade first. See the Migration Guide for the full list of breaking changes. Loading a Model Loading a multi-vector model looks exactly like loading any other Sentence Transformers model: from sentence_transformers import MultiVectorEncoder model = MultiVectorEncoder( "lightonai/LateOn" ) To find models that work, look for the multi-vector and sentence-transformers tags on the Hub. Any model with those tags loads with the line above, whether it started life as a PyLate checkpoint, a Stanford-NLP ColBERT checkpoint, or a ColPali-family model for visual document retrieval. We're working through the ecosystem to get that tag onto every model that works, so the list keeps growing. Underneath, MultiVectorEncoder reads each of the formats these checkpoints have been published in over the years, so PyLate and Stanford-NLP checkpoints load directly even where the tag hasn't been added yet: from sentence_transformers import MultiVectorEncoder # Native Sentence Transformers checkpoints. PyLate builds on the same schema, # so any PyLate checkpoint loads identically model = MultiVectorEncoder( "lightonai/LateOn" ) model = MultiVectorEncoder( "mixedbread-ai/mxbai-edge-colbert-v0-17m" ) model = MultiVectorEncoder( "LiquidAI/LFM2.5-ColBERT-350M" , trust_remote_code= True ) # Any Stanford-NLP ColBERT checkpoint, detected via the `HF_ColBERT` architecture # marker. The inline projection weight and the recipe come from `artifact.metadata` model = MultiVectorEncoder( "colbert-ir/colbertv2.0" ) model = MultiVectorEncoder( "answerdotai/answerai-colbert-small-v1" ) # A bare transformer: a fresh random projection is appended, so training is required model = MultiVectorEncoder( "answerdotai/ModernBERT-base" ) Visual document retrieval models are the exception. ColPali-family checkpoints ship in colpali-engine's own format, which carries no information Sentence Transformers can use, so each one needs a small configuration added to its repository before it loads. Most of that work is done and waiting to be merged. See Supported Models for the current state and how to load them today. Inspecting What a Checkpoint Configured Multi-vector models carry a handful of recipe knobs that differ per checkpoint: marker prefixes for queries and documents, length caps, whether queries are padded out with [MASK] tokens, and which tokens are skipped when scoring documents. All of them live in the module configs, so print(model) shows you exactly what you loaded. Here's the original ColBERTv2 checkpoint, which pads every query to exactly 32 tokens and truncates documents at 180: from sentence_transformers import MultiVectorEncoder model = MultiVectorEncoder( "colbert-ir/colbertv2.0" ) print (model) """ MultiVectorEncoder( (0): Transformer({..., 'document_length': 180, 'query_expansion': {'strategy': 'fixed', 'attend': False, 'token': None, 'length': 32}}) (1): Dense({'in_features': 768, 'out_features': 128, 'bias': False, ...}) (2): MultiVectorMask({'skiplist_words': ['!', '"', '#', ...], 'skiplist_tasks': ['document'], ...}) (3): Normalize({...}) ) """ print (model.prompts) # {'query': '[unused0] ', 'document': '[unused1] '} That's the classic ColBERT pipeline: a Transformer producing contextualized token embeddings, a token-level Dense projecting each of them to 128 dimensions, a MultiVectorMask deciding which tokens count during scoring, and a token-level Normalize . Other checkpoints fill in different values. lightonai/GTE-ModernColBERT-v1 uses the same four modules with [Q] and [D] prompts, no query expansion, and caps of 48 and 300. You rarely need to touch any of this, since every released checkpoint configures its own. It matters when you build a model from a bare backbone, which is covered in Creating Custom Models . One value is worth checking against your own data, though. document_length truncates, so anything past it never reaches the index. For example, a 662-token passage through LateOn's cap of 300 comes back as 273 vectors, with the rest of the passage simply gone. Most of these checkpoints were trained on short passages, so if your chunks are longer than the cap, you can lift it for a single call with encode_document(..., processing_kwargs={"text": {"max_length": 512}}) , keeping in mind that you would be running the model past the length it was trained on and that the index grows roughly in proportion. Multi-vector models tend to tolerate that well. On MLDR , a long-document retrieval benchmark, the multilingual siblings of the pair above show the gap clearly: mLateOn scores 77.92 against mDenseOn's 51.59 . Encoding Queries and Documents Multi-vector models are asymmetric, so queries and documents go through different prefixes, different length caps, and different scoring masks. Unlike many dense models, where the two are interchangeable, encode_query() and encode_document() are required to get correct embeddings: from sentence_transformers import MultiVectorEncoder model = MultiVectorEncoder( "lightonai/mLateOn" ) queries = [ "What is the capital of France?" ] documents = [ "Paris is the capital of France." , "Berlin is the capital and largest city of Germany, by both area and population." , ] query_embeddings = model.encode_query(queries) document_embeddings = model.encode_document(documents) print (query_embeddings[ 0 ].shape) # (10, 128) print (document_embeddings[ 0 ].shape, document_embeddings[ 1 ].shape) # (10, 128) (19, 128) What you get back is a list of 2D tensors, one per input, each of shape (num_tokens, embedding_dim) . Unlike dense embeddings, you can't stack these into one rectangular tensor, because every input has its own token count. The second document is longer than the first, so it comes back as a taller matrix. Each call applies the model's own recipe for you. encode_query prepends the query marker, expands the query to a fixed length if the checkpoint asks for it, and caps it at the query length. encode_document prepends the document marker, caps at the document length, and drops any skiplisted tokens (punctuation, for most checkpoints) from the scoring mask. The usual encode() arguments all still apply, so batch_size , show_progress_bar , convert_to_numpy , device , and multi-process pools work the way you'd expect: document_embeddings = model.encode_document( documents, batch_size= 64 , show_progress_bar= True , ) Scoring with MaxSim model.similarity() computes the full all-pairs MaxSim matrix: from sentence_transformers import MultiVectorEncoder model = MultiVectorEncoder( "lightonai/LateOn" ) query_embeddings = model.encode_query([ "Which planet is known as the Red Planet?" ]) document_embeddings = model.encode_document([ "Venus is often called Earth's twin because of its similar size and proximity." , "Mars, known for its reddish appearance, is often referred to as the Red Planet." , "Jupiter, the largest planet in our solar system, has a prominent red spot." , "Saturn, famous for its rings, is sometimes mistaken for the Red Planet." , ]) scores = model.similarity(query_embeddings, document_embeddings) print (scores) # tensor([[10.7942, 11.1104, 10.9743, 11.0811]]) Mars wins, as it should, though the runners-up are close behind. Saturn also contains the literal phrase "the Red Planet", and Jupiter is a planet with a red spot, so a token-level operator has plenty to latch onto in all three. The ordering is what matters. Scores often sit this close together, as GLInt shows by measuring the spread across a full candidate pool. MaxSim takes a maximum per query token, so a document will usually give every query token some decent best match, and scores start from a floor. Contextualized token embeddings are also anisotropic, clustering in a narrow cone rather than spreading out, so even arbitrary token pairs tend to score high. There is also model.similarity_pairwise() , for when you already have matched pairs and just want the pair scores instead of the full similarity matrix: scores = model.similarity_pairwise(query_embeddings, document_embeddings[: 1 ]) print (scores) # tensor([10.7942]) Score Magnitude and MeanMaxSim MaxSim sums over query tokens, so its magnitude scales with how many query tokens there are, which means you can't compare scores across models with different query recipes. LateOn encodes the Red Planet query above as 12 tokens. Run that same query and those same documents through ColBERTv2, which pads and truncates every query to exactly 32 tokens, and the scores land in a completely different range: model = MultiVectorEncoder( "colbert-ir/colbertv2.0" ) # ... same encode_query / encode_document / similarity calls ... print (scores) # tensor([[12.7970, 27.1945, 23.8495, 24.5656]]) Within one model the ordering is all you need, but if you want scores on a bounded scale, switch the model's similarity function to MeanMaxSim, which divides by the query token count. Back on LateOn: model = MultiVectorEncoder( "lightonai/LateOn" , similarity_fn_name= "meanmaxsim" ) # or on an already-loaded model: model.similarity_fn_name = "meanmaxsim" print (model.similarity(query_embeddings, document_embeddings)) # tensor([[0.8995, 0.9259, 0.9145, 0.9234]]) Now every score is an average cosine similarity in [-1, 1] , although you'll only see [0, 1] in practice. Semantic Search If your corpus is small, exhaustive MaxSim over all of it is the simplest thing that works. Encode the corpus once, then score each query against everything: import time from datasets import load_dataset from sentence_transformers import MultiVectorEncoder dataset = load_dataset( "sentence-transformers/natural-questions" , split= "train[:5000]" ) # Several questions share an answer passage, so drop repeats but keep the order corpus = list ( dict .fromkeys(dataset[ "answer" ])) # 5,000 rows -> 4,874 passages model = MultiVectorEncoder( "lightonai/LateOn" ) corpus_embeddings = model.encode_document(corpus, show_progress_bar= True ) query = "when did richmond last play in a preliminary final" start = time.perf_counter() query_embeddings = model.encode_query([query]) scores = model.similarity(query_embeddings, corpus_embeddings)[ 0 ] # 98ms top_scores, top_indices = scores.topk( 3 ) print ( f"Search took {(time.perf_counter() - start) * 1000 : .1 f} ms" ) for score, index in zip (top_scores.tolist(), top_indices.tolist()): print ( f" {score: .4 f} {corpus[index][: 100 ]} " ) """ Search took 122.7ms 11.9192 Richmond Football Club Richmond began 2017 with 5 straight wins, a feat it had not achieved 11.7591 2017 AFL Grand Final The 2017 AFL Grand Final was an Australian rules football game contest 11.6710 Battle of Appomattox Court House The Battle of Appomattox Court House (Virginia, U.S.), fou """ Those 4,874 passages encoded in 20 seconds on an RTX 3090, and each search takes about 120ms end to end, most of that the MaxSim scoring against all 608,414 token vectors. This is exact, but it scales linearly in total corpus tokens and keeps every token vector in memory, so reach for it when you have a few thousand documents rather than a few million. The runnable version of this script is semantic_search.py . Past that size you want a real late-interaction index, which Sentence Transformers doesn't ship. It doesn't need to, because these indexes store whatever encode_document produced, so you encode here and hand the token embeddings to something built for them. Indexing has working snippets for four of the options, and the section directly below covers how to skip the index entirely. Retrieve and Rerank You can also get late-interaction quality without maintaining a late-interaction index, by using a multi-vector model as your reranker . A fast bi-encoder narrows a large corpus to a handful of candidates, then the multi-vector model rescores only those: from datasets import load_dataset from sentence_transformers import MultiVectorEncoder, SentenceTransformer from sentence_transformers.util import semantic_search dataset = load_dataset( "sentence-transformers/natural-questions" , split= "train[:50000]" ) corpus = list ( dict .fromkeys(dataset[ "answer" ])) retriever = SentenceTransformer( "jinaai/jina-embeddings-v5-text-nano-retrieval" ) reranker = MultiVectorEncoder( "perplexity-ai/pplx-embed-v1-late-0.6b" , trust_remote_code= True ) # First stage: index the corpus once with a fast bi-encoder corpus_embeddings = retriever.encode_document(corpus, convert_to_tensor= True , show_progress_bar= True ) # Retrieve the top 50 query = "when did richmond last play in a preliminary final" hits = semantic_search(retriever.encode_query([query], convert_to_tensor= True ), corpus_embeddings, top_k= 50 )[ 0 ] candidates = [corpus[hit[ "corpus_id" ]] for hit in hits] # Second stage: rescore just those candidates with MaxSim query_embeddings = reranker.encode_query([query]) document_embeddings = reranker.encode_document(candidates) scores = reranker.similarity(query_embeddings, document_embeddings)[ 0 ] for index in scores.argsort(descending= True )[: 3 ].tolist(): print ( f" {scores[index].item(): .4 f} {candidates[index][: 100 ]} " ) Only the 50 candidates are ever encoded as multi-vectors, so your index stays a normal dense index and the token vectors are transient. This is the same role a cross-encoder plays in a retrieve-and-rerank stack, but a multi-vector model is considerably cheaper per candidate. You encode the documents in one batch and score them with a matrix multiplication, instead of one forward pass per query-document pair. The runnable script is retrieve_rerank.py , which prints the timings of both stages. Indexing Several vector databases index and score multi-vectors natively: Qdrant since v1.10, Weaviate since v1.29, Vespa for years now, LanceDB since v0.15.0, and VectorChord , which adds a MaxSim operator to Postgres that plain pgvector doesn't have. Milvus joined them in v2.6.4, under array-of-structs rather than the unrelated feature it calls multi-vector search. If you would rather not run a server at all, LightOn's fast-plaid is a pip install away and implements PLAID directly, and PyLate wraps it in a fuller retrieval stack. A few others get you partway. OpenSearch and Elasticsearch can rescore candidates with MaxSim but not retrieve on it, and the Elasticsearch field is additionally in technical preview and Enterprise-tier. turbopuffer has late-interaction indexing in private beta. The snippets below index text, but nothing in them is text-specific. encode_document hands back the same list of token-vector matrices whether the document was a passage, a page image, an audio clip, or a video, so the ColPali-style models from Visual Document Retrieval go into any of these unchanged. There are simply more vectors per document, which is what makes Token Pooling worth reaching for sooner there. fast-plaid, Qdrant, Weaviate, and Vespa all take exactly what encode_document returns, so the code is the same up to the client library. Here's a working snippet for each, run against the 4,874 passages and 608,414 token vectors from the Semantic Search example. Each one carries the ingestion and query times it produced on one machine (RTX 3090, i7-13700K), with no tuning beyond what the code shows, to give a sense of the shape of the work. All four answer the query faster than the 98ms model.similarity took in that section, and three of them do it on the CPU, since fast-plaid is the only one here using the GPU. All four returned the same three passages in the same order as the exhaustive PyTorch MaxSim earlier in this post, and the three databases reproduce its scores to four decimals! That is because their snippets score every document, which is affordable at this size and removes approximation as a variable. fast-plaid is approximate by design, so its scores differ slightly. The notes under each one say what changes when you switch to an approximate index, which is where rankings start to drift. fast-plaid fast-plaid is LightOn's Rust implementation of PLAID, the index ColBERT was originally built around. There's no server to start, and it reads the tensors encode_document hands back without any conversion. # pip install sentence-transformers datasets fast-plaid from datasets import load_dataset from fast_plaid import search from sentence_transformers import MultiVectorEncoder dataset = load_dataset( "sentence-transformers/natural-questions" , split= "train[:5000]" ) corpus = list ( dict .fromkeys(dataset[ "answer" ])) model = MultiVectorEncoder( "lightonai/LateOn" ) query = "when did richmond last play in a preliminary final" document_embeddings = model.encode_document(corpus, batch_size= 32 ) query_embedding = model.encode_query(query) fast_plaid = search.FastPlaid(index= "natural-questions" , device= "cuda" ) # 4,874 documents (608,414 token vectors) indexed in 5s fast_plaid.create(documents_embeddings=document_embeddings) results = fast_plaid.search(queries_embeddings=query_embedding.unsqueeze( 0 ), top_k= 3 ) # 11ms for index, score in results[ 0 ]: print ( f" {score: .4 f} {corpus[index][: 90 ]} " ) """ 11.8828 Richmond Football Club Richmond began 2017 with 5 straight wins, a feat it had not achieve 11.7676 2017 AFL Grand Final The 2017 AFL Grand Final was an Australian rules football game contes 11.6758 Battle of Appomattox Court House The Battle of Appomattox Court House (Virginia, U.S.), fo """ The index argument is a directory, not just a label, so the index is written to disk as it is built. Pointing a new FastPlaid at the same path reopens it for searching or for adding more documents, instead of rebuilding from the embeddings each time. On this corpus it occupies 92 MB, against 311.5 MB for the raw float32 vectors. This is the only one of the four that is approximate, and it is the one place in this section where the scores do not match the exhaustive MaxSim. PLAID prunes with centroids and stores quantized residuals, so the three scores drift by a few hundredths in both directions against the 11.9192 / 11.7591 / 11.6710 computed earlier. The ranking is unaffected here, and that is the trade PLAID is making. It was designed for corpora far larger than this one, where scanning everything is not an option. Qdrant Qdrant needs a server: docker run -p 6333:6333 qdrant/qdrant . The client also has a local mode ( QdrantClient(":memory:") ) that needs no server, but it's a pure-Python reimplementation, so use it for trying things out rather than for timing them. # pip install sentence-transformers datasets qdrant-client from datasets import load_dataset from qdrant_client import QdrantClient, models from sentence_transformers import MultiVectorEncoder dataset = load_dataset( "sentence-transformers/natural-questions" , split= "train[:5000]" ) corpus = list ( dict .fromkeys(dataset[ "answer" ])) model = MultiVectorEncoder( "lightonai/LateOn" ) query = "when did richmond last play in a preliminary final" document_embeddings = model.encode_document(corpus, batch_size= 32 ) query_embedding = model.encode_query(query) client = QdrantClient( "http://localhost:6333" ) client.create_collection( collection_name= "natural-questions" , vectors_config=models.VectorParams( size=model.get_embedding_dimension(), distance=models.Distance.COSINE, multivector_config=models.MultiVectorConfig( comparator=models.MultiVectorComparator.MAX_SIM ), # MaxSim never walks the HNSW graph, so skip building one hnsw_config=models.HnswConfigDiff(m= 0 ), ), ) # 4,874 documents (608,414 token vectors) ingested in 26.3s client.upload_points( collection_name= "natural-questions" , points=[ models.PointStruct( id =idx, vector=embedding, payload={ "text" : text}) for idx, (embedding, text) in enumerate ( zip (document_embeddings, corpus)) ], batch_size= 64 , ) results = client.query_points( collection_name= "natural-questions" , query=query_embedding, limit= 3 , with_payload= True , ).points # 18ms for result in results: print ( f" {result.score: .4 f} {result.payload[ 'text' ][: 90 ]} " ) """ 11.9192 Richmond Football Club Richmond began 2017 with 5 straight wins, a feat it had not achieve 11.7591 2017 AFL Grand Final The 2017 AFL Grand Final was an Australian rules football game contes 11.6710 Battle of Appomattox Court House The Battle of Appomattox Court House (Virginia, U.S.), fo """ MAX_SIM is the only comparator Qdrant offers, and hnsw_config=HnswConfigDiff(m=0) is their recommendation for late-interaction fields, since the vectors are used for rescoring rather than graph traversal. Note that Qdrant themselves suggest reserving late interaction for reranking a few hundred candidates rather than scanning a whole collection, which is the Retrieve and Rerank pattern. At 4,874 documents the full scan costs 18ms and is exact, but that doesn't extrapolate. Weaviate Weaviate needs a server too: docker run -p 8080:8080 -p 50051:50051 cr.weaviate.io/semitechnologies/weaviate:1.34.0 . Multi-vector support needs 1.29 or newer, and the embedded mode isn't available on Windows. # pip install sentence-transformers datasets weaviate-client import weaviate from datasets import load_dataset from sentence_transformers import MultiVectorEncoder from weaviate.classes.config import Configure, DataType, Property from weaviate.classes.query import MetadataQuery dataset = load_dataset( "sentence-transformers/natural-questions" , split= "train[:5000]" ) corpus = list ( dict .fromkeys(dataset[ "answer" ])) model = MultiVectorEncoder( "lightonai/LateOn" ) query = "when did richmond last play in a preliminary final" document_embeddings = model.encode_document(corpus, batch_size= 32 ) query_embedding = model.encode_query(query) client = weaviate.connect_to_local() collection = client.collections.create( "Documents" , # self_provided turns on MaxSim late interaction vector_config=[Configure.MultiVectors.self_provided(name= "colbert" )], properties=[Property(name= "text" , data_type=DataType.TEXT)], ) # 4,874 documents (608,414 token vectors) ingested in 41s with collection.batch.fixed_size(batch_size= 64 ) as batch: for text, embedding in zip (corpus, document_embeddings): batch.add_object(properties={ "text" : text}, vector={ "colbert" : embedding.tolist()}) results = collection.query.near_vector( near_vector=query_embedding.tolist(), target_vector= "colbert" , limit= 3 , return_metadata=MetadataQuery(distance= True ), ) # 17ms for result in results.objects: # Weaviate reports the MaxSim score as a negated distance print ( f" {-result.metadata.distance: .4 f} {result.properties[ 'text' ][: 90 ]} " ) """ 11.9192 Richmond Football Club Richmond began 2017 with 5 straight wins, a feat it had not achieve 11.7591 2017 AFL Grand Final The 2017 AFL Grand Final was an Australian rules football game contes 11.6710 Battle of Appomattox Court House The Battle of Appomattox Court House (Virginia, U.S.), fo """ client.close() Defaults are enough here. Weaviate's dynamic ef resolves to 100 for a top-3 query, and this ranking is already exact from about 32 upward. That margin is a property of the embeddings rather than of Weaviate, so it's worth confirming on your own model instead of assuming the defaults hold. Weaviate also supports MUVERA encoding, which made ingestion 3x faster and queries 1.8x faster in our test. It cost far more accuracy than that speed is worth at this size though, since the correct third passage didn't appear even in its top 50. Vespa Vespa also runs in a container, but pyvespa starts it for you, so there's no separate docker run . # pip install sentence-transformers datasets pyvespa from datasets import load_dataset from sentence_transformers import MultiVectorEncoder from vespa.deployment import VespaDocker from vespa.package import ( ApplicationPackage, Document, Field, FirstPhaseRanking, Function, RankProfile, Schema, ) dataset = load_dataset( "sentence-transformers/natural-questions" , split= "train[:5000]" ) corpus = list ( dict .fromkeys(dataset[ "answer" ])) model = MultiVectorEncoder( "lightonai/LateOn" ) query = "when did richmond last play in a preliminary final" document_embeddings = model.encode_document(corpus, batch_size= 32 ) query_embedding = model.encode_query(query) # "dt" is a mapped dimension over the variable token count, "x" the dense 128-dim vector package = ApplicationPackage( name= "colbert" , schema=[ Schema( name= "doc" , document=Document(fields=[ Field(name= "text" , type = "string" , indexing=[ "summary" ]), Field(name= "colbert" , type = "tensor<float>(dt{}, x[128])" , indexing=[ "attribute" ]), ]), rank_profiles=[ RankProfile( name= "colbert" , inputs=[( "query(qt)" , "tensor<float>(qt{}, x[128])" )], functions=[Function( name= "max_sim" , # per query token take the best document token, then sum expression= "sum(reduce(sum(query(qt) * attribute(colbert), x), max, dt), qt)" , )], first_phase=FirstPhaseRanking(expression= "max_sim" ), ) ], ) ], ) app = VespaDocker(port= 8080 ).deploy(application_package=package) # ~40s to boot # Vespa reads a mixed tensor as {token index: vector}, for documents and queries alike def to_tensor ( embedding ): return { str (token): vector for token, vector in enumerate (embedding.tolist())} # 4,874 documents (608,414 token vectors) ingested in ~80s app.feed_iterable( ({ "id" : str (idx), "fields" : { "text" : text, "colbert" : to_tensor(embedding)}} for idx, (text, embedding) in enumerate ( zip (corpus, document_embeddings))), schema= "doc" , ) response = app.query(body={ "yql" : "select text from doc where true" , "ranking.profile" : "colbert" , "hits" : 3 , "input.query(qt)" : to_tensor(query_embedding), }) # ~75ms warm, ~115ms on the first call for hit in response.hits: print ( f" {hit[ 'relevance' ]: .4 f} {hit[ 'fields' ][ 'text' ][: 90 ]} " ) """ 11.9192 Richmond Football Club Richmond began 2017 with 5 straight wins, a feat it had not achieve 11.7591 2017 AFL Grand Final The 2017 AFL Grand Final was an Australian rules football game contes 11.6710 Battle of Appomattox Court House The Battle of Appomattox Court House (Virginia, U.S.), fo """ Vespa asks for the most upfront structure of the four, because you're declaring a ranking pipeline rather than just an index. In exchange you get to write MaxSim out as a tensor expression and see exactly what it computes. This version puts MaxSim in first-phase over where true , which scores all 4,874 documents and is why the output matches exhaustive MaxSim exactly. It's deliberately not what Vespa recommends at scale. Their ColBERT sample app stores int8-binarized vectors and moves MaxSim into second-phase to rerank a cheaper first stage. Moving to that phased setup needs care, because second-phase rescores only the best 100 candidates by default, and here that window left two of the three correct passages unscored entirely. Raising rerank-count to cover your candidate set fixes that, though at this size the phased version still came out slower than simply scanning everything. Visual Document Retrieval Late interaction is the state of the art for visual document retrieval, matching a text query against page images , with charts, tables, and layout intact, and no OCR step. This is what the ColPali family of models does, and those checkpoints load and run through the same API, with the revision pinning the open pull request that adds this one's Sentence Transformers configuration ( Supported Models has the full list). Image documents are passed as URLs, local paths, or PIL images: from sentence_transformers import MultiVectorEncoder model = MultiVectorEncoder( "vidore/colqwen2.5-v0.2" ) queries = [ "What is the variable represented on the y-axis of the graph?" , "Total outlay is maximum in which year?" , ] images = [ "https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/doc1.jpg" , "https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/doc2.jpg" , "https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/doc3.jpg" , "https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/doc4.jpg" , ] query_embeddings = model.encode_query(queries) document_embeddings = model.encode_document(images) print (query_embeddings[ 0 ].shape, document_embeddings[ 0 ].shape) # (25, 128) (755, 128) scores = model.similarity(query_embeddings, document_embeddings) print (scores) # tensor([[13.8672, 12.3115, 12.1670, 11.0293], # [ 7.2012, 14.7207, 6.9414, 6.9746]]) Each query retrieves its own page (the diagonal), and the second query separates much more cleanly than the first, since only one of the four pages is about outlay over time. The code is unchanged. Underneath, the processor handles the visual prompt and the image patches, and MaxSim scores query text tokens against document image patches. A page holds many separate regions, which is exactly what makes late interaction a natural fit here, since a single vector would have to average a chart, a table, and three paragraphs into one summary. That fidelity costs index space, though. The shapes above are 755 token vectors for one page against 25 for the query, where a Natural Questions passage from earlier averaged about 125, so token pooling is worth reaching for earlier here than it is for text. These are VLMs, so plan for the memory they need. The table in Supported Models runs from 252M to 8.8B parameters, and the small end of it stays practical on CPU where the multi-billion ones don't. Page images are the common case, but they're not the only non-text modality. Sentence Transformers accepts text, images, audio, and video, and a checkpoint supports whichever of those its processor does, which model.modalities reports. A single document can combine modalities too, by passing a dict like {"text": ..., "image": ...} in place of a bare value. Multimodal Embedding & Reranker Models covers multimodal models in Sentence Transformers more broadly, and the Usage documentation lists exactly which input formats each modality accepts. Audio Retrieval vidore/colqwen-omni-v0.1 is built on Qwen2.5-Omni and takes all four modalities. Retrieving a recorded conversation with it is the same two calls as retrieving a page: # pip install -U "sentence-transformers[audio,video]" import torch from datasets import Audio, load_dataset from sentence_transformers import MultiVectorEncoder model = MultiVectorEncoder( "vidore/colqwen-omni-v0.1" , model_kwargs={ "dtype" : torch.bfloat16}, ) print (model.modalities) # ['text', 'image', 'audio', 'video', 'message'] # 20 recorded conversations, averaging 28 seconds each dataset = load_dataset( "eustlb/dailytalk-conversations-grouped" , split= "train[:20]" ) dataset = dataset.cast_column( "audio" , Audio(sampling_rate= 16_000 )) audio = [row[ "array" ] for row in dataset[ "audio" ]] # raw mono waveforms, float32 at 16 kHz query_embeddings = model.encode_query([ "medicine for car nausea" ]) document_embeddings = model.encode_document(audio, batch_size= 2 ) scores = model.similarity(query_embeddings, document_embeddings)[ 0 ] top_scores, top_indices = scores.topk( 3 ) for score, index in zip (top_scores.tolist(), top_indices.tolist()): print ( f" {score: .4 f} { ' / ' .join(dataset[index][ 'texts' ][: 2 ])} " ) """ 50.8902 Excuse me? Do you have anything for a carsickness? / Yes, but you look fine. 46.1028 Excuse me, could you tell me where you have got that music book? / Certainly. Let me see. Oh, it's on that shelf. 46.0514 Jeff, I'm going to the supermarket. Do you want to come with me? / I think the supermarket is closed now. """ ColQwen-Omni was trained purely on image-text pairs, so its audio retrieval is zero-shot. It never heard a training example, and there is no transcription step anywhere in the pipeline. The query says nausea where the recording says carsickness , and it still picks the pharmacy conversation out of twenty by a wide margin. Video Retrieval Video works the same way, but sample the frames or it will eat your VRAM. Its release blogpost is blunt about this, that video "is very memory-intensive, so it's best suited for short clips": import torch from sentence_transformers import MultiVectorEncoder model = MultiVectorEncoder( "vidore/colqwen-omni-v0.1" , model_kwargs={ "dtype" : torch.bfloat16}, ) # Sparse, low-resolution frames: 0.5 fps rather than the full frame rate model[ 0 ].processing_kwargs.update( { "video" : { "max_pixels" : 32 * 28 * 28 , "do_sample_frames" : True , "fps" : 0.5 }} ) query_embeddings = model.encode_query([ "How to cook Mapo Tofu?" ]) document_embeddings = model.encode_document([ "https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/mapo_tofu.mp4" , "https://huggingface.co/datasets/sentence-transformers/example-documents/resolve/main/zhajiang_noodle.mp4" , ], batch_size= 1 ) print (model.similarity(query_embeddings, document_embeddings)) # tensor([[53.3100, 51.0561]]) At 1 fps and full resolution the same pair of videos produces 8,426 and 5,137 token vectors and peaks at 20.8 GB of VRAM, against 4,240 and 2,446 vectors and 12.5 GB here, for a model that occupies 9.0 GB on its own. The ranking is identical either way. Long audio wants the same treatment, and the release blogpost recommends 30-second chunks, which come to roughly 800 tokens each. Interpretability Because MaxSim is a sum of per-query-token maxima, a ranking decomposes exactly, so every point of a document's score belongs to one query token and one document token. That lets you answer "why did this rank here?" precisely, rather than by eye. For image documents, sentence_transformers.multi_vector_encoder.interpretability overlays that decomposition onto the page as the standard ColPali heatmap, either aggregated over the query or one map per query token. Asking "How much was spent on water resources and power?" against the outlays page from above, this is where the water token went: heatmap.py is the runnable version, including the masking step that lines the document embedding up with the patch grid. Text documents have no patch grid to overlay, but the same decomposition applies. text_similarity_map.py ranks a corpus and then attributes the top hit's score token by token, here on the Natural Questions corpus from earlier with the 32M-parameter mxbai-edge-colbert-v0-32m : Query: when did richmond last play in a preliminary final Top 3 of 4874 documents by exhaustive MaxSim (191.0ms): 12.3489 Richmond Football Club Richmond began 2017 with 5 straight wins, a feat it had not achieved since 19 12.1771 2017 AFL Grand Final The 2017 AFL Grand Final was an Australian rules football game contested betwee 12.0591 2018 UEFA Champions League Final The 2018 UEFA Champions League Final was the final match of the 201 query token best document token sim share when since 0.9154 7.4% did had 0.9675 7.8% rich rich 0.9764 7.9% mond mond 0.9856 8.0% last to 0.9249 7.5% play game 0.9384 7.6% in the 0.9732 7.9% a a 0.9587 7.8% preliminary preliminary 0.9394 7.6% final final 0.9654 7.8% -------------------------------------------------------- 3 special tokens 2.8038 22.7% MaxSim score 12.3489 100.0% rich , mond , preliminary , and final matched themselves, while when settled on since and play on game . Three special tokens contribute 22.7% of the score while carrying none of the query's content. Below this table the script prints the passage itself, with the winning tokens highlighted in place. Token Pooling If the index footprint worries you, the most effective knob is to store fewer token vectors. HierarchicalTokenPooling implements the token pooling technique from Clavié, Chaffin, and Adams by clustering each document's token vectors with Ward linkage on cosine distance and replacing each cluster with its mean, keeping roughly 1 / pool_factor of the tokens. Within one document a lot of token vectors end up close to each other, so much of what you drop is redundancy rather than signal: from datasets import load_dataset from sentence_transformers import MultiVectorEncoder from sentence_transformers.multi_vector_encoder.modules import HierarchicalTokenPooling dataset = load_dataset( "sentence-transformers/natural-questions" , split= "train[:5000]" ) documents = list ( dict .fromkeys(dataset[ "answer" ])) model = MultiVectorEncoder( "lightonai/LateOn" ) pooling = HierarchicalTokenPooling(pool_factor= 2 ) document_embeddings = model.encode_document(documents, token_pooling=pooling) There are three places to apply it, depending on when you want to pay for it: # 1. Per encode call, as above document_embeddings = model.encode_document(documents, token_pooling=pooling) # 2. Standalone, on embeddings you already have saved (e.g. list of [num_tokens, num_dims] tensors) pooled = pooling.pool(document_embeddings) # 3. Baked into the model, so every consumer of the checkpoint gets pooled documents model.append(HierarchicalTokenPooling(pool_factor= 2 )) model.save_pretrained( "my-pooled-colbert" ) By default, pooling applies to documents only, since queries are short and are the side you can't afford to distort. On the Natural Questions corpus from earlier, the reduction tracks pool_factor closely, and pooling all 608k token vectors took about 6 seconds: pool_factor Token vectors Reduction float32 index 1 (off) 608,414 1.00x 311.5 MB 2 305,438 1.99x 156.4 MB 3 204,407 2.98x 104.7 MB 4 153,936 3.95x 78.8 MB A cluster mean is a worse match for a query token than the best of its members was, and the coarser the clusters, the more that shows. The original experiments measured that cost on BEIR and found very little of it, at 100.6% of the unpooled retrieval performance on average at pool_factor=2 , and 99.0% at pool_factor=3 . Halving your index for free is a good deal, so 2 is a reasonable place to start. How much it costs on your data is corpus-specific though, so measure it with an evaluator before you settle on a factor. The runnable comparison is token_pooling.py . How far you can push pool_factor is also partly a property of the model. LightOn's hierarchical pooling regularization trains for exactly that, shaping the embedding space so pooling costs less and reporting 99.4% retention at 5x compression. Training with that regularizer isn't in Sentence Transformers yet, but the resulting checkpoints are ordinary PyLate models, so lightonai/LateOn-hpool-regularized loads and pools like any other. Speeding Up Inference Multi-vector models run through the same backend machinery as the rest of Sentence Transformers, so you get torch (default), onnx , and openvino , alongside half precision, Flash Attention, and torch.compile . On GPU, fp16 with Flash Attention is the best configuration we measured, at 2.44x the throughput of fp32 with no measurable retrieval quality loss. Flash Attention helps multi-vector models more than most, because documents are only truncated and never padded to a shared length, so your batches have widely varying sequence lengths that unpadding can exploit: from sentence_transformers import MultiVectorEncoder model = MultiVectorEncoder( "lightonai/GTE-ModernColBERT-v1" , model_kwargs={ "attn_implementation" : "flash_attention_2" , "dtype" : "float16" }, ) Models with non-attend query expansion ( attend=False , which covers the Stanford-NLP checkpoints like colbert-ir/colbertv2.0 and answerdotai/answerai-colbert-small-v1 ) reject Flash Attention at load time. Flash Attention strips attention_mask=0 positions, so the [MASK] expansion tokens that MaxSim scores would never receive an attention update. Use "sdpa" for those models. On CPU, OpenVINO is your better bet where the architecture is supported, and int8 quantization buys a further speedup at a cost of about 0.4% accuracy. See Speeding up Inference for the full benchmark details, the export and quantization helpers, and a flowchart for picking a backend. Evaluating a Model MultiVectorNanoBEIREvaluator runs the NanoBEIR suite of 13 small BEIR subsets with MaxSim scoring, and needs no data preparation on your side: from sentence_transformers import MultiVectorEncoder from sentence_transformers.multi_vector_encoder.evaluation import MultiVectorNanoBEIREvaluator model = MultiVectorEncoder( "lightonai/GTE-ModernColBERT-v1" ) evaluator = MultiVectorNanoBEIREvaluator(batch_size= 16 ) results = evaluator(model) print ( f" {evaluator.primary_metric} : {results[evaluator.primary_metric]: .4 f} " ) This also makes it easy to check the claim from the top of this post. lightonai/LateOn and lightonai/DenseOn were trained by LightOn on the same data with the same ModernBERT backbone and the same 149M parameters, differing only in whether they keep one vector per token or pool down to one per document. Running both over all 13 NanoBEIR datasets isolates what that choice buys: NanoBEIR dataset LateOn (multi-vector, 128d) DenseOn (dense, 768d) MSMARCO 0.7194 0.6517 NQ 0.7810 0.7511 HotpotQA 0.9295 0.8802 FEVER 0.9702 0.9612 ClimateFEVER 0.4887 0.4846 DBPedia 0.6836 0.6748 QuoraRetrieval 0.9795 0.9687 Touche2020 0.5938 0.5673 ArguAna 0.5562 0.5660 NFCorpus 0.3949 0.3851 SciFact 0.7978 0.8057 SCIDOCS 0.4469 0.4484 FiQA2018 0.5871 0.6491 Mean 0.6868 0.6764 Late interaction wins on 9 of the 13 datasets and on the mean, by roughly one NDCG point. The four it loses (ArguAna, FiQA2018, SCIDOCS, and SciFact) are the shape of the tradeoff you should expect, a real gain in retrieval quality at the same model size, paid for in index footprint, rather than a universal win on every dataset. The same pair scores 57.22 against 56.20 on the full 15-dataset BEIR, a comparable gap, so the margin is not an artifact of the small benchmark. Alongside NanoBEIR, MultiVectorInformationRetrievalEvaluator , MultiVectorRerankingEvaluator , MultiVectorTripletEvaluator , and MultiVectorDistillationEvaluator cover the usual evaluation setups on your own data. They're documented in the Evaluation API Reference . Coming from PyLate or colpali-engine MultiVectorEncoder absorbs the modeling, inference, training, and evaluation of both libraries. Every PyLate checkpoint loads directly, and Supported Models lists the colpali-engine checkpoints along with the revision to pass where one is still needed. If you're migrating, these are the calls that change: PyLate Sentence Transformers pylate.models.ColBERT(model_name_or_path=...) MultiVectorEncoder(...) model.encode(..., is_query=True) model.encode_query(...) model.encode(..., is_query=False) model.encode_document(...) pylate.scores.colbert_scores model.similarity pylate.indexes.PLAID / pylate.retrieve.ColBERT no equivalent, keep PyLate's PLAID or see Indexing colpali-engine Sentence Transformers ColQwen2.from_pretrained(...) + ColQwen2Processor MultiVectorEncoder(...) processor.process_queries(...) + model(**batch) model.encode_query(queries) processor.process_images(...) + model(**batch) model.encode_document(images) processor.score_multi_vector(qs, ds) model.similarity(query_embeddings, document_embeddings) mask_non_image_embeddings=True MultiVectorMask(keep_only_token_ids=[...]) HierarchicalTokenPooler HierarchicalTokenPooling colpali_engine.interpretability sentence_transformers.multi_vector_encoder.interpretability The difference worth knowing about is on bare (non-ColBERT) checkpoints, where PyLate's ColBERT("bert-base-uncased") applies the classic recipe by default while MultiVectorEncoder("bert-base-uncased") builds a plain stack, leaving the prefixes, query expansion, and skiplist as explicit choices. The training loss and evaluator equivalents, and the data-handling differences, are in the Migration Guide . Save compatibility is one-way in every case. PyLate, Stanford-NLP ColBERT, and colpali-engine checkpoints all load into MultiVectorEncoder , but MultiVectorEncoder.save_pretrained output isn't loadable by any of them. Supported Models Models carrying the multi-vector and sentence-transformers tags on the Hub are the list that stays current, and we're working to get those tags onto every model that works. The tables below are what we test against directly, so treat them as a starting point rather than the full set. For text retrieval in particular, any PyLate or Stanford-NLP ColBERT checkpoint loads whether or not it carries the tag yet. Some entries need a small Sentence Transformers configuration added to their repository first, and several of those are still open pull requests at the time of writing. Where a revision is listed below, pass it until that pull request is merged, after which the plain model name is enough: model = MultiVectorEncoder( "vidore/colqwen-omni-v0.1" , revision= "refs/pr/N" ) Text Retrieval Models These load with their trained prefix tokens, query expansion, and punctuation skiplist recovered from the saved configuration. The NanoBEIR column reports the mean NDCG@10 (higher is better) across the 13 NanoBEIR datasets , each a 50-query subsample of a BEIR dataset, as a fast proxy for English text retrieval quality. We used the MultiVectorNanoBEIREvaluator to compute the scores for the primarily-English models. A - means the model was not evaluated on it. Note that NanoBEIR is a small benchmark, and its scores aren't a substitute for evaluating on your own data, which is always the right way to pick a model. Model Parameters Dimensionality NanoBEIR Notes lightonai/LateOn-regularized 149M 128 0.6897 - lightonai/LateOn-hpool-regularized 149M 128 0.6876 - lightonai/LateOn 149M 128 0.6868 - LiquidAI/LFM2.5-ColBERT-350M 353M 128 0.6864 needs trust_remote_code=True lightonai/mLateOn 307M 128 0.6851 - VAGOsolutions/SauerkrautLM-Multi-ModernColBERT 149M 128 0.6741 - lightonai/GTE-ModernColBERT-v1 149M 128 0.6720 - topk-io/Iso-ModernColBERT 149M 128 0.6687 - perplexity-ai/pplx-embed-v1-late-0.6b 596M 128 0.6662 needs trust_remote_code=True VAGOsolutions/SauerkrautLM-Multi-Reason-ModernColBERT 149M 128 0.6616 - lightonai/ColBERT-Zero 149M 128 0.6569 - answerdotai/answerai-colbert-small-v1 33M 96 0.6550 - mixedbread-ai/mxbai-edge-colbert-v0-32m 32M 64 0.6524 - jinaai/jina-colbert-v2 559M 128 0.6517 needs trust_remote_code=True DataScience-UIBK/Reason-mxbai-colbert-v0.1-32m 32M 128 0.6486 - LiquidAI/LFM2-ColBERT-350M 353M 128 0.6441 - lightonai/ColBERT-Zero-unsupervised-noprompts 149M 128 0.6430 - mixedbread-ai/mxbai-edge-colbert-v0-17m 17M 48 0.6407 - lightonai/colbertv2.0 110M 128 0.6201 - lightonai/LateOn-Code 149M 128 0.6169 - lightonai/Agent-ModernColBERT 149M 128 0.6164 - lightonai/Reason-ModernColBERT 149M 128 0.6078 - colbert-ir/colbertv2.0 110M 128 0.6053 - VAGOsolutions/SauerkrautLM-Reason-EuroColBERT 212M 128 0.6039 - VAGOsolutions/SauerkrautLM-EuroColBERT 212M 128 0.5965 - antoinelouis/colbert-xm 853M 128 0.5915 - mixedbread-ai/mxbai-colbert-large-v1 335M 128 0.5733 - lightonai/LateOn-Code-edge 17M 48 0.5274 - NeuML/biomedbert-base-colbert 110M 128 0.4320 - NeuML/colbert-bert-tiny 4M 128 0.4035 - yjoonjang/colbert-ko-v1 149M 128 - - ytu-ce-cosmos/turkish-colbert 111M 256 - - samheym/GerColBERT 110M 128 - - bclavie/JaColBERT 111M 128 - - bclavie/JaColBERTv2 111M 128 - - answerdotai/JaColBERTv2.4 111M 128 - revision="refs/pr/2" answerdotai/JaColBERTv2.5 111M 128 - revision="refs/pr/5" Visual Document Retrieval Models ColPali-style models embed page images as documents and text as queries. The NanoViDoRe column reports the mean NDCG@10 (higher is better) across NanoViDoRe v3 , a compact visual document retrieval benchmark spanning 8 subsets (computer science, energy, finance in English and French, HR, industrial, pharmaceuticals, and physics). Like with NanoBEIR, NanoViDoRe is a small benchmark which shouldn't replace evaluation on your own data. Model Parameters Dimensionality NanoViDoRe Notes webAI-Official/webAI-ColVec1.1-8b 8.4B 640 0.6580 needs trust_remote_code=True webAI-Official/webAI-ColVec1.1-4b 4.5B 640 0.6520 needs trust_remote_code=True vultr/VultronRetrieverPrime-Qwen3.5-8B 8.39B 320 0.6423 revision="refs/pr/2" vultr/VultronRetrieverCore-Qwen3.5-4.5B 4.54B 320 0.6410 revision="refs/pr/1" tencent/EVIE-Preview-4.5B 4.54B 128 0.6405 - nvidia/nemotron-colembed-vl-8b-v2 8.77B 4096 0.6374 revision="refs/pr/4" , needs trust_remote_code=True athrael-soju/colqwen3.5-4.5B-v3 4.54B 320 0.6358 - TomoroAI/tomoro-colqwen3-embed-8b 8.8B 320 0.6206 needs trust_remote_code=True nvidia/nemotron-colembed-vl-4b-v2 4.83B 2560 0.6200 revision="refs/pr/7" , needs trust_remote_code=True OpenSearch-AI/Ops-Colqwen3-4B 4.44B 2560 0.6150 revision="refs/pr/5" , needs trust_remote_code=True nvidia/llama-nemotron-colembed-vl-3b-v2 4.41B 3072 0.6112 revision="refs/pr/3" , needs trust_remote_code=True tsystems/colqwen2.5-3b-multilingual-v1.0 3.75B 128 0.6039 revision="refs/pr/3" tsystems/colqwen2.5-3b-multilingual-v1.0-merged 3.75B 128 0.6027 revision="refs/pr/1" TomoroAI/tomoro-colqwen3-embed-4b 4.4B 320 0.6019 needs trust_remote_code=True nomic-ai/colnomic-embed-multimodal-7b 8.29B 128 0.5942 revision="refs/pr/3" nomic-ai/colnomic-embed-multimodal-3b 3.75B 128 0.5929 revision="refs/pr/6" VAGOsolutions/SauerkrautLM-ColQwen3-8b-v0.1 8.15B 128 0.5819 revision="refs/pr/1" Metric-AI/ColQwen2.5-3b-multilingual-v1.0 3.75B 128 0.5763 revision="refs/pr/2" vultr/VultronRetrieverFlash-Qwen3.5-0.8B 853M 320 0.5693 revision="refs/pr/2" VAGOsolutions/SauerkrautLM-ColQwen3-4b-v0.1 4.44B 128 0.5656 revision="refs/pr/1" VAGOsolutions/SauerkrautLM-ColQwen3-2b-v0.1 2.13B 128 0.5530 revision="refs/pr/1" Verm1ion/ColTurk-VDR-Qwen3VL-4B-v1.0 4.44B 320 0.5434 revision="refs/pr/1" vidore/colqwen2.5-v0.2 3.8B 128 0.5402 - vidore/colqwen2.5-v0.1 3.8B 128 0.5395 - vidore/colqwen-omni-v0.1 4.4B 128 0.5309 - VAGOsolutions/SauerkrautLM-ColQwen3-1.7b-Turbo-v0.1 1.76B 128 0.5035 revision="refs/pr/1" vidore/colpali-v1.3 2.9B 128 0.4802 - vidore/colpali-v1.3-hf 2.9B 128 0.4793 - vidore/colpali-v1.2 2.9B 128 0.4691 - vidore/colqwen2-v1.0 2.2B 128 0.4685 - vidore/colqwen2-v0.1 2.2B 128 0.4526 - vidore/colpali 2.9B 128 0.4516 - vidore/colpali-v1.1 2.9B 128 0.4314 - VAGOsolutions/SauerkrautLM-ColLFM2-450M-v0.1 451M 128 0.4249 revision="refs/pr/1" vidore/colsmolvlm-v0.1 2.1B 128 0.4054 - VAGOsolutions/SauerkrautLM-ColMinistral3-3b-v0.1 4.25B 128 0.3978 revision="refs/pr/1" vidore/colpali-hard-v1.1 2.9B 128 0.3949 - vidore/colSmol-500M 507M 128 0.3459 - vidore/colSmol-256M 256M 128 0.2673 - ModernVBERT/colmodernvbert 252M 128 0.2632 - vidore/colpali-v1.2-hf 2.9B 128 - - vidore/colqwen2-v1.0-hf 2.2B 128 - - Most of these are LoRA adapter repositories, with the adapter applied directly onto its base at load time. Some also have a -merged sibling on the Hub (e.g. vidore/colpali-v1.3-merged ) with the adapter already folded into the weights. The three -hf entries are the transformers-native *ForRetrieval ports. They load without any configuration, but use more modeling from transformers and less from sentence_transformers . Generally, it's preferable to use the original models instead, as the ports score approximately the same. Acknowledgements Late interaction in Sentence Transformers rests on a lot of earlier work. Thanks to Omar Khattab and Matei Zaharia for ColBERT , which everything here descends from, and to the LightOn team (Antoine Chaffin, Raphael Sourty, Paulo Moura, and Amélie Chatelain) for PyLate and fast-plaid , which carried late interaction for years and shaped a good deal of the API described above. Thanks to the ColPali team (Manuel Faysse, Hugues Sibille, Tony Wu, Bilel Omrani, Gautier Viaud, Céline Hudelot, and Pierre Colombo) for ColPali and colpali-engine, which brought late interaction to page images, and to Benjamin Clavié, Antoine Chaffin, and Griffin Adams for token pooling . Thanks as well to the core MTEB team, Kenneth Enevoldsen and Roman Solomatin among many others, for MTEB and for the kind of hidden work that keeps information retrieval research running. And thanks to everyone who trained and released the checkpoints in Supported Models . Without them this post would have had nothing to measure. Additional Resources Documentation Multi-Vector Encoder > Usage Multi-Vector Encoder > Pretrained Models Multi-Vector Encoder > Creating Custom Models Multi-Vector Encoder > Speeding up Inference Multi-Vector Encoder > API Reference Installation Migration Guide Example Scripts Semantic Search Retrieve and Rerank Token Pooling ColPali Heatmaps Text Similarity Maps NanoBEIR Evaluation Training To learn how to train or finetune these models on your own data: See the companion blogpost: Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers . Multi-Vector Encoder > Training Overview Multi-Vector Encoder > Loss Overview Multi-Vector Encoder > Training Examples LateOn and mLateOn training scripts : LightOn's PyLate recipes for LateOn, mLateOn, DenseOn, and mDenseOn, where the finetuning scripts show practical details like splitting a 16,384-example batch into mini-batches of 16. Hugging Face Hub Multi-vector models on the Hub Sentence Transformers datasets on the Hub Companion Blogposts Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers : the direct training companion to this post. Training and Finetuning Embedding Models with Sentence Transformers : the general training guide for text-only dense embedding models. Training and Finetuning Reranker Models with Sentence Transformers : Cross Encoder training, the other way to add a precise second stage. Training and Finetuning Sparse Embedding Models with Sentence Transformers : SPLADE and other sparse encoders, which combine well with late interaction in hybrid search. Multimodal Embedding & Reranker Models with Sentence Transformers : single-vector multimodal models, the dense counterpart to ColPali-style retrieval. Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers : includes a Visual Document Retrieval walkthrough with single-vector models. 🪆 Introduction to Matryoshka Embedding Models : shrink dense embeddings by dimension, the way token pooling shrinks multi-vector ones by count.