ResearchBenchmarksOpen-ended Dialogue on OOD AverageFollow60.5Win RateCausalRM30.44438.24746.0553.853Jan 29, 2026Evaluation ResultsMethodMethodLinksWin RateTie RateLoss RateCausalRMOpponent=SFTOpponent=SFT2026.0160.528.910.6CausalRMOpponent=Standard RMOpponent=Standard RM2026.0151.332.616.1CausalRMOpponent=InfoRMOpponent=InfoRM2026.0138.741.819.5CausalRMOpponent=GoalRMOpponent=GoalRM2026.0131.650.617.8