Nemotron ColEmbed V2: Top-Performing Late Interaction embedding models for Visual Document Retrieval

• URL: http://arxiv.org/abs/2602.03992v1
• Date: Tue, 03 Feb 2026 20:26:44 GMT
• Title: Nemotron ColEmbed V2: Top-Performing Late Interaction embedding models for Visual Document Retrieval
• Authors: Gabriel de Souza P. Moreira, Ronay Ak, Mengyao Xu, Oliver Holworthy, Benedikt Schifferer, Zhiding Yu, Yauhen Babakhin, Radek Osmulski, Jiarui Cai, Ryan Chesler, Bo Liu, Even Oldridge,
• Abstract summary: Motivated by the growing demand for visual document retrieval, we introduce Nemotron ColEmbed V2, a family of models that achieve state-of-the-art performance on the ViDoRe benchmarks.<n>We describe the main techniques used across data processing, training, and post-training to build our top-performing models.
• Score: 19.23621110865551
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Retrieval-Augmented Generation (RAG) systems have been popular for generative applications, powering language models by injecting external knowledge. Companies have been trying to leverage their large catalog of documents (e.g. PDFs, presentation slides) in such RAG pipelines, whose first step is the retrieval component. Dense retrieval has been a popular approach, where embedding models are used to generate a dense representation of the user query that is closer to relevant content embeddings. More recently, VLM-based embedding models have become popular for visual document retrieval, as they preserve visual information and simplify the indexing pipeline compared to OCR text extraction. Motivated by the growing demand for visual document retrieval, we introduce Nemotron ColEmbed V2, a family of models that achieve state-of-the-art performance on the ViDoRe benchmarks. We release three variants - with 3B, 4B, and 8B parameters - based on pre-trained VLMs: NVIDIA Eagle 2 with Llama 3.2 3B backbone, Qwen3-VL-4B-Instruct and Qwen3-VL-8B-Instruct, respectively. The 8B model ranks first on the ViDoRe V3 leaderboard as of February 03, 2026, achieving an average NDCG@10 of 63.42. We describe the main techniques used across data processing, training, and post-training - such as cluster-based sampling, hard-negative mining, bidirectional attention, late interaction, and model merging - that helped us build our top-performing models. We also discuss compute and storage engineering challenges posed by the late interaction mechanism and present experiments on how to balance accuracy and storage with lower dimension embeddings.

Related papers

• Towards Mixed-Modal Retrieval for Universal Retrieval-Augmented Generation [72.34977512403643]
  Retrieval-Augmented Generation (RAG) has emerged as a powerful paradigm for enhancing large language models (LLMs) by retrieving relevant documents from an external corpus.<n>Existing RAG systems primarily focus on unimodal text documents, and often fall short in real-world scenarios where both queries and documents may contain mixed modalities (such as text and images)<n>We propose Nyx, a unified mixed-modal to mixed-modal retriever tailored for Universal Retrieval-Augmented Generation scenarios.
  arXiv  Detail & Related papers  (2025-10-20T09:56:43Z)
• Leveraging Generative Models for Real-Time Query-Driven Text Summarization in Large-Scale Web Search [54.987957691350665]
  Query-Driven Text Summarization (QDTS) aims to generate concise and informative summaries from textual documents based on a given query.<n>Traditional extractive summarization models, based primarily on ranking candidate summary segments, have been the dominant approach in industrial applications.<n>We propose a novel framework to pioneer the application of generative models to address real-time QDTS in industrial web search.
  arXiv  Detail & Related papers  (2025-08-28T08:51:51Z)
• SitEmb-v1.5: Improved Context-Aware Dense Retrieval for Semantic Association and Long Story Comprehension [77.93156509994994]
  We show how to represent short chunks in a way that is conditioned on a broader context window to enhance retrieval performance.<n>Existing embedding models are not well-equipped to encode such situated context effectively.<n>Our method substantially outperforms state-of-the-art embedding models.
  arXiv  Detail & Related papers  (2025-08-03T23:59:31Z)
• Llama Nemoretriever Colembed: Top-Performing Text-Image Retrieval Model [20.055106781946417]
  llama-nemoretriever-colembed is a unified text-image retrieval model that delivers state-of-the-art performance across multiple benchmarks.<n>The 3B model achieves state of the art performance, scoring NDCG@5 91.0 on ViDoRe V1 and 63.5 on ViDoRe V2, placing first on both leaderboards as of June 27, 2025.
  arXiv  Detail & Related papers  (2025-07-07T22:20:04Z)
• Revisiting Birds Eye View Perception Models with Frozen Foundation Models: DINOv2 and Metric3Dv2 [6.42131197643513]
  We introduce an innovative application of Metric3Dv2's depth information as a PseudoLiDAR point cloud incorporated into the Simple-BEV architecture.<n>This integration results in a +3 IoU improvement compared to the Camera-only model.
  arXiv  Detail & Related papers  (2025-01-14T13:51:14Z)
• VisRAG: Vision-based Retrieval-augmented Generation on Multi-modality Documents [66.42579289213941]
  Retrieval-augmented generation (RAG) is an effective technique that enables large language models to utilize external knowledge sources for generation.<n>We introduce VisRAG, which tackles this issue by establishing a vision-language model (VLM)-based RAG pipeline.<n>In this pipeline, instead of first parsing the document to obtain text, the document is directly embedded using a VLM as an image and then retrieved to enhance the generation of a VLM.
  arXiv  Detail & Related papers  (2024-10-14T15:04:18Z)
• NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models [38.41524186248607]
  We introduce NV-Embed, incorporating architectural designs, training procedures, and curated datasets.<n>For model architecture, we propose a latent attention layer to obtain pooled embeddings.<n>For training algorithm, we introduce a two-stage contrastive instruction-tuning method.
  arXiv  Detail & Related papers  (2024-05-27T17:59:45Z)
• SPRINT: A Unified Toolkit for Evaluating and Demystifying Zero-shot Neural Sparse Retrieval [92.27387459751309]
  We provide SPRINT, a unified Python toolkit for evaluating neural sparse retrieval. We establish strong and reproducible zero-shot sparse retrieval baselines across the well-acknowledged benchmark, BEIR. We show that SPLADEv2 produces sparse representations with a majority of tokens outside of the original query and document.
  arXiv  Detail & Related papers  (2023-07-19T22:48:02Z)
• DSI++: Updating Transformer Memory with New Documents [95.70264288158766]
  We introduce DSI++, a continual learning challenge for DSI to incrementally index new documents. We show that continual indexing of new documents leads to considerable forgetting of previously indexed documents. We introduce a generative memory to sample pseudo-queries for documents and supplement them during continual indexing to prevent forgetting for the retrieval task.
  arXiv  Detail & Related papers  (2022-12-19T18:59:34Z)
• Incorporating Relevance Feedback for Information-Seeking Retrieval using Few-Shot Document Re-Ranking [56.80065604034095]
  We introduce a kNN approach that re-ranks documents based on their similarity with the query and the documents the user considers relevant. To evaluate our different integration strategies, we transform four existing information retrieval datasets into the relevance feedback scenario.
  arXiv  Detail & Related papers  (2022-10-19T16:19:37Z)
• SPLADE v2: Sparse Lexical and Expansion Model for Information Retrieval [11.38022203865326]
  SPLADE model provides highly sparse representations and competitive results with respect to state-of-the-art dense and sparse approaches. We modify the pooling mechanism, benchmark a model solely based on document expansion, and introduce models trained with distillation. Overall, SPLADE is considerably improved with more than $9$% gains on NDCG@10 on TREC DL 2019, leading to state-of-the-art results on the BEIR benchmark.
  arXiv  Detail & Related papers  (2021-09-21T10:43:42Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.