ResearchBenchmarksOpen-ended Dialogue on TruthfulQA OODFollow52.2WinsCausalRM30.8836.41541.9547.485Jan 29, 2026Evaluation ResultsMethodMethodLinksWinsTiesLossesCausalRMOpponent=SFTOpponent=SFT2026.0152.241.26.6CausalRMOpponent=Standard RMOpponent=Standard RM2026.0150.930.418.7CausalRMOpponent=InfoRMOpponent=InfoRM2026.0144.139.716.2CausalRMOpponent=GoalRMOpponent=GoalRM2026.0131.752.915.4