Improve Nano dataset README provenance and overview

README.md

@@ -511,7 +511,9 @@ tags:
 
 This dataset is a Nano-style retrieval dataset for [HAKARI-bench](https://github.com/hakari-bench/hakari-bench).
 
-NanoRTEB contains English and code-heavy retrieval splits derived from publicly loadable RTEB retrieval tasks.
+NanoRTEB samples open English RTEB tasks for compact evaluation across practical domain search settings, including legal case and statute retrieval, finance question answering, healthcare and medical question answering, and code-oriented retrieval.
+
+It keeps the Nano query-corpus-qrels layout while providing BM25, dense `harrier_oss_v1_270m`, and `reranking_hybrid` candidate subsets for first-stage retrieval and reranking diagnostics.
 
 ## Usage
 
@@ -601,6 +603,8 @@ Dense means `microsoft/harrier-oss-v1-270m` with the `web_search_query` prompt a
 
 ## Source Links
 
+- Source benchmark: `RTEB`
+- [RTEB benchmark announcement](https://huggingface.co/blog/rteb)
 - [MTEB project](https://github.com/embeddings-benchmark/mteb)
 
 ## License