sentence-transformers
An open-source Python framework for text vectorization that encodes sentences and paragraphs into fixed-dimensional semantic vectors. It is widely used in NLP tasks such as semantic search, retrieval-augmented generation (RAG), and text similarity computation.
Timeline (last 90 days)
EmbeddingGemma 2's context window was expanded from v1's 2048 tokens to 8192, MTEB Code improved from 68.76 to 78.68, and it already supports Ollama, llama.cpp, sentence-transformers, MLX, vLLM, and browser-side transformers.js; there is currently no independent retrieval benchmark test for v2, and the SciFact score of 0.7861 belongs to v1.
Lightweight embedding models such as sentence-transformers (on the order of hundreds of millions of parameters) already achieve throughput on CPUs sufficient for daily incremental indexing needs, without requiring a dedicated GPU.
Version 1.2.3 upgrades the sentence-transformers dependency from 5.2.3 to 5.6.0.
Multiple Negatives Ranking Loss (MNRL) treats each question's corresponding passage as a positive example within a batch and uses all other passages in the batch as negative examples, eliminating the need to manually collect negative samples.
If the same passage appears twice in the same batch, the loss function will treat the duplicate passage as a wrong answer; a no-duplicate sampler should be used to avoid this.
sentence-transformers can be run locally, making them particularly suitable for privacy-sensitive scenarios.
Commonly used embedding models include OpenAI's text-embedding-ada-002 and the locally runnable sentence-transformers.
Nils Reimers is best known for the sentence-transformers open-source library and the Sentence-BERT (SBERT) series of models behind it
The project adopts a RAG approach, using sentence-transformers to generate vectors and FAISS for vector retrieval, enabling semantic search over the company's data table schemas, KPI definitions, and business rules.
Common semantic search implementation solutions include OpenAI's text-embedding model series or open-source sentence-transformers, paired with vector databases such as Pinecone and Qdrant.
2 more timeline events
All Facts (13)
OpenAI's text-embedding-ada-002 and the open-source sentence-transformers series are commonly used embedding models.
60%UnverifiedEmbeddingGemma 2's context window was expanded from v1's 2048 tokens to 8192, MTEB Code improved from 68.76 to 78.68, and it already supports Ollama, llama.cpp, sentence-transformers, MLX, vLLM, and browser-side transformers.js; there is currently no independent retrieval benchmark test for v2, and the SciFact score of 0.7861 belongs to v1.
50%UnverifiedLightweight embedding models such as sentence-transformers (on the order of hundreds of millions of parameters) already achieve throughput on CPUs sufficient for daily incremental indexing needs, without requiring a dedicated GPU.
50%UnverifiedVersion 1.2.3 upgrades the sentence-transformers dependency from 5.2.3 to 5.6.0.
50%UnverifiedIf the same passage appears twice in the same batch, the loss function will treat the duplicate passage as a wrong answer; a no-duplicate sampler should be used to avoid this.
50%UnverifiedMultiple Negatives Ranking Loss (MNRL) treats each question's corresponding passage as a positive example within a batch and uses all other passages in the batch as negative examples, eliminating the need to manually collect negative samples.
50%Unverifiedsentence-transformers can be run locally, making them particularly suitable for privacy-sensitive scenarios.
50%UnverifiedCommonly used embedding models include OpenAI's text-embedding-ada-002 and the locally runnable sentence-transformers.
50%UnverifiedNils Reimers is best known for the sentence-transformers open-source library and the Sentence-BERT (SBERT) series of models behind it
50%UnverifiedThe project adopts a RAG approach, using sentence-transformers to generate vectors and FAISS for vector retrieval, enabling semantic search over the company's data table schemas, KPI definitions, and business rules.
50%UnverifiedCommon semantic search implementation solutions include OpenAI's text-embedding model series or open-source sentence-transformers, paired with vector databases such as Pinecone and Qdrant.
50%UnverifiedUsing lightweight embedding models such as sentence-transformers' MiniLM requires tens of milliseconds per inference, while large embedding models can reach hundreds of milliseconds.
50%UnverifiedLocally runnable sentence-transformers models (such as all-MiniLM-L6-v2) can work entirely offline, whereas cloud-based models such as OpenAI's text-embedding-3-small offer higher accuracy but require an internet connection.
50%