ResearchBenchmarksOpen-ended Dialogue on SHP OODFollow77.5Win RateCausalRM25.91639.30852.766.092Jan 29, 2026Evaluation ResultsMethodMethodLinksWin RateTie RateLoss RateCausalRMOpponent=SFTOpponent=SFT2026.0177.511.810.7CausalRMOpponent=Standard RMOpponent=Standard RM2026.015526.318.7CausalRMOpponent=InfoRMOpponent=InfoRM2026.0133.845.620.6CausalRMOpponent=GoalRMOpponent=GoalRM2026.0127.950.821.3