Found interesting approach (instead of traditional embeddings) to store and provide extra knowledge to the LLM for generating response (RAG). So called “knowledge maps”. As far as I understand, they can provide higher quality in accuracy of augmental retrieval. How every there are neither details about their structure, nor explanation how llm interacts with them. Maybe somebody can share more details?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Knowledge Base in realtime api | 3 | 283 | January 22, 2025 | |
| What do you think of GraphRAG? | 0 | 1056 | May 23, 2024 | |
| How can RAG systems be improved for more complex queries | 3 | 3429 | October 31, 2023 | |
| Strategy to Train Model using R.A.G | 5 | 2234 | October 18, 2023 | |
| How real is graph RAG, I see we are gradually evolving to GRAPH RAG | 1 | 1153 | May 11, 2024 |