Atlas · GenAI 2026
Embedding Models
Custom embedding model fine-tuning (BGE, Nomic)
conceptPeak: 2024EmbeddingsAI consensus: 3/3
Prerequisites
- hardNLP
Fine-tuning embedding models requires understanding how embeddings represent semantic relationships and what contrastive learning optimizes
- mediumLLM Fine-Tuning
Embedding fine-tuning uses similar training loop concepts (learning rate, epochs, validation) as SFT — familiarity with fine-tuning accelerates learning
Recommended reference
Xiao et al. (2024) 'C-Pack: Packaged Resources to Advance General Chinese Embedding' (BGE paper) — best practices for embedding fine-tuning
Notes from AI deep research
Anthropic Opus
BGE, Nomic, E5. Fine-tuning pod branzowy slownik = ogromny boost jakosci RAG
OpenAI Deep Research
Dobór modeli, normalizacja, dystanse [OA#31]
Google Deep Think
Dostrajanie pod słownik branżowy [G#46]
Related skills
- → is part of: Retrieval-Augmented Generation(3/3)
- → is part of: Semantic Search(3/3)
- ← is an instance of: Sentence-Transformers(0/3)