Update README.md

README.md

@@ -54,7 +54,7 @@ pipeline_tag: sentence-similarity
     - [How can I reduce overall inference cost ?](#how-can-i-reduce-overall-inference-cost)
     - [How do I reduce vector storage cost?](#how-do-i-reduce-vector-storage-cost)
     - [How do I offer hybrid search to improve accuracy?](#how-do-i-offer-hybrid-search-to-improve-accuracy)
-    - [CMTEB numbers](#cmteb-numbers)
+- [MTEB numbers](#mteb-numbers)
 - [Roadmap](#roadmap)
 - [Notes on Reproducing:](#notes-on-reproducing)
 - [Reference:](#reference)
@@ -177,11 +177,17 @@ The below numbers are with mDPR model, but miniDense_arabic_v1 should give a eve
 
 *Note: MIRACL paper shows a different (higher) value for BM25 Arabic, So we are taking that value from BGE-M3 paper, rest all are form the MIRACL paper.*
 
-#### MTEB numbers:
+# MTEB numbers:
 MTEB is a general purpose embedding evaluation benchmark covering wide range of tasks, but miniDense models (like BGE-M3) are predominantly tuned for retireval tasks aimed at search & IR based usecases.
 So it makes sense to evaluate our models in retrieval slice of the MTEB benchmark.
 
-##### Long Document Retrieval 
+#### MIRACL Retrieval
+
+Refer tables above 
+
+#### Long Document Retrieval 
+
+This is very ambitious eval because we have not trained for long context, the max_len was 512 for all the models below.
 
 <center>
 <img src="./ar_metrics_4.png" width=150%/>
@@ -189,9 +195,10 @@ So it makes sense to evaluate our models in retrieval slice of the MTEB benchmar
 </center>
 
 
-##### X-lingual Retrieval 
+#### X-lingual Retrieval 
 
-Almost all models below are monolingual arabic models so they have no notion of any other languages. But the below table shows how our model excels in cross-lingual scenarios.
+Almost all models below are monolingual arabic models so they have no notion of any other languages. But the below table shows how our model excels in cross-lingual scenarios owing to its deep multilingual understanding.
+This also explains its competitive performance when compared to models lot larger.
 
 <center>
 <img src="./ar_metrics_5.png" width=80%/>