Arctic-Embed: Scalable, Efficient, and Accurate Text Embedding Models

By Luke Merrick, Danmei Xu, Gaurav Nuti, Daniel Campos. 2024

Subfields: Information Retrieval & Search, Representation & Self-Supervised Learning

Citations in the corpus

• CoRNStack: High-Quality Contrastive Data for Better Code Retrieval and Reranking (2024, unknown): “The encoder is initialized with Arctic-Embed-M (Merrick et al., 2024), a text encoder supporting an extended context length of 8,192 tokens and pretrained on large-scale web query-document pairs, along with public text retrieval datasets (Yang et al., 2018; Kwiatkowski et al., 2019; Thorne et al.,…”
• The TellTail of Embeddings: Fingerprinting Retrievers in Black-Box Systems (2026, unknown): “5 [44]; gtr-t5-base [45]; minilm-l6 and minilm-l12 [46]; mpnet [47]; snowflake-v2 [48]; gemma-300m [49]; bge-base [50]; and the closed-weights model text-embedding-3-small (OpenAI) [51].”
• GigaEmbeddings: Efficient Russian Language Embedding Model (2025, unknown): “Similarly, Merrick et al. (2024) generates synthetic queries while grounding them with real negative documents, observing that LLMs struggle to produce high-quality negatives independently.”
• Fantastic (small) Retrievers and How to Train Them: mxbai-edge-colbert-v0 Tech Report (2025, unknown): “A similar form of this effect was described in the snowflake-arctic embeddings tech report [19], where stratification of training examples by origin dataset yielded superior results.”
• Retrieval Capabilities of Large Language Models Scale with Pretraining FLOPs (2025, unknown): “There has also been a recent uptick in BERT-style embedding models including Nomic Embed (Nussbaum et al., 2024), Arctic-Embed (Merrick et al., 2024), and ModernBERT (Warner et al., 2024).”
• SweRank: Software Issue Localization with Code Ranking (2025, unknown): “The models considered include: Arctic-Embed (Merrick et al., 2024), primarily pre-trained on English text retrieval data; CodeRankEmbed, pre-trained on 22 million NL-to-Code examples (Suresh et al., 2024); and Arctic-Embed-v2.”
• Can't Hide Behind the API: Stealing Black-Box Commercial Embedding Models (2024, unknown): “Inter-estingly, research by Merrick et al. (2024) has shown that embedding models trained from BERT base can achieve competitive performance, with an MTEB Retrieval Average nDCG@10 of 55 .”
• DenseOn with the LateOn: Fully Open Dense and Late-Interaction Models for Multilingual, Long-Context, and Code Search (2026, unknown): “Since pre-training retrieval scores are not always reliable predictors of post-fine-tuning quality (Merrick et al., 2024), each candidate recipe is evaluated after supervised fine-tuning on the full BEIR suite.”
• Finding the Right Tables and Columns: A Benchmark and Corpus-Adaptive Embeddings for SQL Schema Retrieval (2026, unknown): “We fine-tune Arctic-Embed-M with a contrastive objective, using one positive and six hard negatives per query, and mix the in-domain synthetic data with schema-linking pairs derived from the public text-to-SQL training splits of Spider and BIRD (Yu et al., 2018; Li et al., 2023b).”