Nomic Embed Vision: Expanding the Latent Space

By Zach Nussbaum, Brandon Duderstadt, Andriy Mulyar. arXiv.org, 2024

26 citations recorded by the source corpus.

This technical report describes the training of nomic-embed-vision, a highly performant, open-code, open-weights image embedding model that shares the same latent space as nomic-embed-text. Together, nomic-embed-vision and nomic-embed-text form the first unified latent space to achieve high performance across vision, language, and multimodal tasks.

Subfields: Computer Vision and Pattern Recognition, Multimodal Learning, Representation & Self-Supervised Learning

View Paper: https://arxiv.org/abs/2406.18587

Citations in the corpus

• CLeaR: A Unified Framework for Resolving the Leakage-Degradation Dilemma in Style Transfer (2026, unknown): “We add SigLIP [39] and Nomic [19] as additional VFMs.”
• Don't Contrast the Impossible: Region-Constrained Batching for Contrastive User Modeling on a Local Community Platform (2026, unknown): “For item features, we use embeddings obtained from a pretrained embedding model [8, 9], which takes an item’s title, content, metadata and image as input and produces the corresponding item representations.”
• jina-embeddings-v5-omni: Geometry-preserving Embeddings via Locked Aligned Towers (2026, unknown): “We take note of previous work in frozen-tower methods based on the CLIP architecture, such as LiT [39] and Nomic Embed Vision [25], which freeze the text encoder while adapting the other media towers.”
• Revealing Geography-Driven Signals in Zone-Level Claim Frequency Models: An Empirical Study using Environmental and Visual Predictors (2026, unknown): “5 model, a pretrained Vision Transformer optimized for semantic similarity (Nussbaum, Duderstadt, and Mulyar, 2024).”
• HIVE: Query, Hypothesize, Verify An LLM Framework for Multimodal Reasoning-Intensive Retrieval (2026, unknown): “• Nomic Embed Vision [26]: Shares an embedding space between a vision encoder and a strong text model, enabling zero-shot multimodal retrieval.”
• Visual Product Search Benchmark (2026, unknown): “Works such as VLM2Vec [16], Jina Embeddings v4 [12], and Nomic Embed Vision [22] aim to produce embeddings that generalize across modalities, domains, and retrieval tasks without task-specific fine-tuning.”
• Unlocking Multimodal Document Intelligence: From Current Triumphs to Future Frontiers of Visual Document Retrieval (2026, unknown): “Addressing efficiency, ColModernVBERT (Teiletche et al., 2025) demonstrates that a compact, bidirectional language encoder can be effectively aligned with a vision encoder to generate granular embeddings; while MetaEmbed (Xiao et al., 2025d) introduces a fixed set of learnable "Meta Tokens" whose fi”
• Nemotron ColEmbed V2: Top-Performing Late Interaction embedding models for Visual Document Retrieval (2026, unknown): “VLM decoder models have been adapted as contrastive embedding models for visual document retrieval, like Jina CLIP [17] and Nomic Embed Vision [18].”
• MiRAGE: A Multiagent Framework for Generating Multimodal Multihop Question-Answer Dataset for RAG Evaluation (2026, unknown): “The multimodal chunks are embedded with the Nomic model (Nussbaum et al., 2024) for semantic retrieval based on textual queries 3 .”
• Deep Learning Image Embeddings for Ventricular Beat Classification (2025, unknown): “…[19], MobileNet [20], ResNet [21], Wide ResNet [22], ResNeXt [23], ShuffleNet [24], SqueezeNet [25], and the following CLIP architectures: Bootstrapping Language-Image Pre-training [26], DINOv2 [27], CLIP (Contrastive Language–Image Pre-training) [28], MetaCLIP [29], and Nomic-Embed-Vision [30].”
• Semantic topologies in the recursive application of generative AI models (2025, unknown): “…multi-modal (text+image) embedding models exist [15], their text vs image embeddings are not directly comparable (we attempted to do so with [16] but the images were embedded differently enough to the texts that their clusters were entirely disjoint) so we choose to use text-only embedding…”
• Targeted perturbations reveal brain-like local coding axes in robustified, but not standard, ANN-based brain models (2025, unknown): “1 (Iandola et al., 2016), AlexNet (Krizhevsky et al., 2012), CORnet-RT (Kubilius et al., 2018), DenseNet201 (Huang et al., 2017), MobileNet v2 (Sandler et al., 2018)), three self-supervised vision transformers (DINOv2 (Oquab et al., 2023), DreamSim-ViT-B/32 (Fu et al., 2023), Google ViT (Dosovitskiy”
• Colflor: Towards Bert-Size Vision-Language Document Retrieval Models (2025, unknown): “• Contrastive VLMs : Vision-language embedding models trained via contrastive learning in a two-tower setup (text and image encoders), including Jina CLIP [20], Nomic-Vision [21], and SigLIP [22].”
• Solving Inverse Problems in Stochastic Self-Organising Systems through Invariant Representations (2025, unknown): “…of the embedding representations against six other baselines: mean squared error (MSE) for pixel-level comparison, Nomic, a second embedding model [52]; Fréchet Wavelet, a metric based on wavelet packet decomposition [53]; and a set domain-specific metrics: dissimilarity index, which measures…”
• VIBE: Vector Index Benchmark for Embeddings (2025, unknown): “5 (Nussbaum et al., 2024).”