Turned 16 overlapping eval cases into a labeled scam detection set
3.4M tokens
Replaced 16 overlapping eval cases with a labeled set of scams and sketchy-but-legit messages.
Derived from this session's token and cost data. Not shown on the feed.
Comments