Disable native viewer; tighten card prose

README.md

@@ -24,6 +24,7 @@ multilinguality:
 - monolingual
 source_datasets:
 - original
+viewer: false
 ---
 
 # CADGenBench (Inputs)
@@ -101,9 +102,8 @@ One directory per fixture, named by its numeric id. There are two shapes:
 
 ### Data Splits
 
-There is no train/test split — all 81 fixtures are an evaluation set
-(49 generation, 32 editing). The fixture set may grow over time; consumers
-should discover fixtures at runtime rather than hardcoding the list.
+No train/test split — all 81 fixtures form a single evaluation set
+(49 generation, 32 editing).
 
 ## Dataset Creation
 
@@ -123,9 +123,9 @@ evaluation is the only path to a score.
 
 ## Considerations for Using the Data
 
-This is a benchmark **input** set, not a training corpus. Ground truth is not
-included here. To participate, generate one `output.step` per fixture and
-submit through the leaderboard Space.
+This is a benchmark **input** set, not a training corpus. To participate,
+generate one `output.step` per fixture and submit through the leaderboard
+Space.
 
 ## Additional Information