docs: saleforce's sfr rag (#1717)

This PR adds Salesforce's newly released SFR RAG

docs/src/rag/sfr_rag.md

@@ -0,0 +1,17 @@
+**SFR RAG 📑**
+====================================================================
+Salesforce AI Research introduces SFR-RAG, a 9-billion-parameter language model trained with a significant emphasis on reliable, precise, and faithful contextual generation abilities specific to real-world RAG use cases and relevant agentic tasks. They include precise factual knowledge extraction, distinguishing relevant against distracting contexts, citing appropriate sources along with answers, producing complex and multi-hop reasoning over multiple contexts, consistent format following, as well as refraining from hallucination over unanswerable queries.
+
+**[Offical Implementation](https://github.com/SalesforceAIResearch/SFR-RAG)**
+
+<figure markdown="span">
+  ![agent-based-rag](https://raw.githubusercontent.com/lancedb/assets/main/docs/assets/rag/salesforce_contextbench.png)
+  <figcaption>Average Scores in ContextualBench: <a href="https://blog.salesforceairesearch.com/sfr-rag/">Source</a>
+  </figcaption>
+</figure>
+
+To reliably evaluate LLMs in contextual question-answering for RAG, Saleforce introduced [ContextualBench](https://huggingface.co/datasets/Salesforce/ContextualBench?ref=blog.salesforceairesearch.com), featuring 7 benchmarks like [HotpotQA](https://arxiv.org/abs/1809.09600?ref=blog.salesforceairesearch.com) and [2WikiHopQA](https://www.aclweb.org/anthology/2020.coling-main.580/?ref=blog.salesforceairesearch.com) with consistent setups. 
+
+SFR-RAG outperforms GPT-4o, achieving state-of-the-art results in 3 out of 7 benchmarks, and significantly surpasses Command-R+ while using 10 times fewer parameters. It also excels at handling context, even when facts are altered or conflicting.
+
+[Saleforce AI Research Blog](https://blog.salesforceairesearch.com/sfr-rag/)