Upload README.md

README.md

@@ -12,7 +12,7 @@ configs:
   - config_name: default
     data_files:
       - split: train
-        path: final_ai_safety_papers_full_with_metadata_normalized.parquet
+        path: main/*.parquet
 ---
 
 # S2ORC Safety
@@ -31,21 +31,21 @@ It contains `16,806` papers selected through:
 
 ## Main Files
 
-- `final_ai_safety_papers_full_with_metadata_normalized.parquet`
-  - full enriched source rows
+- `main/*.parquet`
+  - sharded full enriched source rows
   - extracted metadata
   - normalized GitHub repo links
   - Hugging Face code mirror links
   - normalized model / dataset / metric / scalar fields
 
-- `paper_metadata.parquet`
-  - metadata extraction only
+- `metadata/*.parquet`
+  - sharded metadata extraction only
 
 - `paper_metadata_summary_normalized.json`
   - corpus-level summary statistics over the normalized metadata fields
 
-- `paper_code_links.parquet`
-  - paper-to-code join table with normalized GitHub URLs and HF mirror paths
+- `code_links/*.parquet`
+  - sharded paper-to-code join table with normalized GitHub URLs and HF mirror paths
 
 ## Contents