Roman Derunets, Siberian Neural Networks (developer-researcher, laboratory of applied digital technologies, NSU)
The team developed a multimodal large language model based on retrieval-augmented generation (RAG): instead of training projection layers or end-to-end training on multimodal data, it "recalls" relevant paragraphs from a text knowledge base given an image or sound.
Continue reading — it's free
Google or a one-time email code. No password.
