Leaderboard default

app.py

@@ -369,6 +369,7 @@ def create_app():
                 - **Total Pairs:** {benchmark_stats.get('total_pairs', 'N/A'):,}
                 - **Categories:** {len(benchmark_stats.get('categories', {}))}
                 - **Avg Hebrew Length:** {benchmark_stats.get('avg_he_length', 0):.0f} chars
+                - **Dataset:** [View on Hugging Face](https://huggingface.co/datasets/{BENCHMARK_DATASET_ID})
                 """)
             
             with gr.Column(scale=1):
@@ -381,7 +382,21 @@ def create_app():
         
         gr.Markdown("---")
         
-        with gr.Tabs(selected=1):  # Default to Leaderboard tab
+        with gr.Tabs(selected=0):  # Default to Leaderboard tab
+            with gr.TabItem("🏆 Leaderboard"):
+                leaderboard_table = gr.Dataframe(
+                    value=format_leaderboard_df(),
+                    label="Model Rankings",
+                    interactive=False,
+                )
+                
+                refresh_btn = gr.Button("🔄 Refresh Leaderboard")
+                
+                comparison_plot = gr.Plot(
+                    value=create_leaderboard_comparison(),
+                    label="Model Comparison"
+                )
+            
             with gr.TabItem("🔬 Evaluate Model"):
                 with gr.Row():
                     with gr.Column(scale=2):
@@ -426,20 +441,6 @@ def create_app():
                         status_text = gr.Markdown("")
                         
                         results_markdown = gr.Markdown("")
-            
-            with gr.TabItem("🏆 Leaderboard"):
-                leaderboard_table = gr.Dataframe(
-                    value=format_leaderboard_df(),
-                    label="Model Rankings",
-                    interactive=False,
-                )
-                
-                refresh_btn = gr.Button("🔄 Refresh Leaderboard")
-                
-                comparison_plot = gr.Plot(
-                    value=create_leaderboard_comparison(),
-                    label="Model Comparison"
-                )
         
         gr.Markdown("""
         ---

dataset/README.md

@@ -25,7 +25,7 @@ A benchmark dataset for evaluating embedding models on Rabbinic Hebrew and Arama
 
 ## Dataset Description
 
-This dataset contains 3,708 parallel text pairs spanning diverse Rabbinic literature across multiple centuries and genres. It is designed for evaluating cross-lingual embedding models on their ability to align Hebrew/Aramaic source texts with English translations.
+This dataset contains parallel text pairs spanning diverse Rabbinic literature across multiple centuries and genres. It is designed for evaluating cross-lingual embedding models on their ability to align Hebrew/Aramaic source texts with English translations.
 
 ### Languages
 
@@ -57,18 +57,10 @@ Each example contains:
 
 ## Intended Use
 
-### Primary Use Case
-
 Evaluating embedding models for cross-lingual retrieval:
 - Given a Hebrew/Aramaic text, can the model find its English translation from a pool of candidates?
 - Models that excel at this task likely capture the semantics of Rabbinic literature well.
 
-### Evaluation Metrics
-
-- **Recall@k**: Percentage of queries where correct translation is in top k results
-- **MRR**: Mean Reciprocal Rank
-- **Bitext Accuracy**: True pair vs random pair classification
-
 ## Source
 
 All texts and translations are from [Sefaria](https://www.sefaria.org), a free library of Jewish texts.
@@ -88,7 +80,7 @@ If you use this dataset, please cite Sefaria:
 @misc{sefaria,
   title = {Sefaria: A Living Library of Jewish Texts},
   url = {https://www.sefaria.org},
-  year = {2024}
+  year = {2026}
 }
 ```