GitHub - FlagOpen/FlagEmbedding: Retrieval and Retrieval-augmented LLMs
GitHub - FlagOpen/FlagEmbedding: Retrieval and Retrieval-augmented LLMshttps://github.com/FlagOpen/FlagEmbedding

GitHub - FlagOpen/FlagEmbedding: Retrieval and Retrieval-augmented LLMs

🚀Discover FlagEmbedding on GitHub! This cutting-edge tool offers retrieval-augmented LLMs like LM-Cocktail & BGE-M3, supporting multi-vector retrieval in 100+ languages. Enhance your models with Activation Beacon & BGE Reranker for accurate document ranking. #AI #GitHub

  • FlagEmbedding focuses on retrieval-augmented LLMs, such as Long-Context LLM and LM-Cocktail.
  • BGE-M3 is a new embedding model supporting dense, lexical, and multi-vector retrieval across 100+ languages and inputs up to 8192 tokens.
  • Activation Beacon extends context length of LLMs effectively and efficiently.
  • LM-Cocktail merges fine-tuned models to enhance performance without losing general capabilities.
  • LLM Embedder augments large language models by supporting various retrieval needs.
  • BGE Reranker uses cross-encoder models for accurate re-ranking of documents.
  • BGE Embedding includes general embedding models fine-tuned for enhanced retrieval abilities.
  • C-MTEB serves as a benchmark for Chinese text embedding.
  • FlagEmbedding projects include Activation-Beacon, LM-Cocktail, LLM-Embedder, BGE Reranker, BGE Embedding, and C-MTEB.
  • Contributors to FlagEmbedding include individuals like Shitao Xiao, Peitian Zhang, and Zheng Liu.