Base Ari

Held-out reward rose from 0.188 to 0.228 across 8 training runs

Best so farNo improvementNot evaluated
Base Ari00.10.20.30.40.50.60.7Held-out eval reward12345678Training runBaseline 0.188BaselineMove checkFEN board16kLonger runSFT→GRPOSmoke testSFT stage