Datasets:

Surpem
/

WEBMemo

@@ -207,7 +207,6 @@ The generator for WEBMemo follows a few strict constraints:
 WEBMemo is synthetic. That makes it useful for controllable fine-tuning, but it also means:
-- the reasoning traces are generated rather than collected from natural human workflows
 - realism depends on fit to your downstream use case
 - chain-of-thought style supervision may not match every deployment policy
 - prompt diversity does not automatically imply benchmark-level difficulty

 WEBMemo is synthetic. That makes it useful for controllable fine-tuning, but it also means:
 - realism depends on fit to your downstream use case
 - chain-of-thought style supervision may not match every deployment policy
 - prompt diversity does not automatically imply benchmark-level difficulty