E5
A series of text embedding models developed by Microsoft, trained through weakly supervised contrastive pre-training followed by supervised fine-tuning, achieving strong performance on multilingual text retrieval tasks.
Core Facts
Timeline (last 90 days)
Embedding models are neural network models that map discrete inputs into dense vectors, with representative examples including OpenAI's text-embedding-ada-002, Voyage AI's voyage series, and open-source BGE, E5, GTE and others.
Even in modern embedding models optimized through instruction fine-tuning or contrastive learning (such as BGE and E5), the anisotropy problem has not been completely eliminated, only reduced in degree.
Agents running in Agent 365 require an assigned license, such as E5 and the Frontier license for autopilot
Mainstream text embedding models are typically trained based on contrastive learning, mapping semantically similar text pairs to nearby positions in vector space and pushing unrelated text pairs further apart.
E5 由微软研究院推出，采用对比学习训练范式
FastEmbed-rs 支持多种主流开源嵌入模型，包括 BGE、E5、all-MiniLM 等
BGE和E5系列在MTEB基准测试上表现出色，其中BGE-M3支持多语言、多粒度和多功能检索
All Facts (8)
OpenAI's text-embedding-ada-002 can map text into a high-dimensional vector space, where semantically similar texts are closer together in the vector space.
65%UnverifiedEmbedding models are neural network models that map discrete inputs into dense vectors, with representative examples including OpenAI's text-embedding-ada-002, Voyage AI's voyage series, and open-source BGE, E5, GTE and others.
50%UnverifiedEven in modern embedding models optimized through instruction fine-tuning or contrastive learning (such as BGE and E5), the anisotropy problem has not been completely eliminated, only reduced in degree.
50%UnverifiedAgents running in Agent 365 require an assigned license, such as E5 and the Frontier license for autopilot
50%UnverifiedMainstream text embedding models are typically trained based on contrastive learning, mapping semantically similar text pairs to nearby positions in vector space and pushing unrelated text pairs further apart.
50%UnverifiedE5 由微软研究院推出，采用对比学习训练范式
50%UnverifiedFastEmbed-rs 支持多种主流开源嵌入模型，包括 BGE、E5、all-MiniLM 等
50%UnverifiedBGE和E5系列在MTEB基准测试上表现出色，其中BGE-M3支持多语言、多粒度和多功能检索
50%