List your company

Logical Reasoning on ZebraLogic (held-out)

71.7AccuracyCORE
48.40454.45260.566.548May 27, 2026

Evaluation results

MethodLinks
COREtraining samples=102026.0571.7
GEPAtraining samples=1002026.0570.7
COREtraining samples=52026.0570
MemRLtraining samples=52026.0568.3
COREtraining samples=1002026.0566.3
GEPAtraining samples=52026.0559.7
MemRLtraining samples=1002026.0558.7
GEPAtraining samples=102026.0557
MemRLtraining samples=102026.0554.3
Episodic RAGtraining samples=52026.0554
GRPOtraining samples=102026.0553.3
GRPOtraining samples=52026.0552.3
GRPOtraining samples=1002026.0552
No Learningtraining samples=52026.0550.9
No Learningtraining samples=102026.0550.9
No Learningtraining samples=1002026.0550.9
Episodic RAGtraining samples=1002026.0549.7
Episodic RAGtraining samples=102026.0549.3