ResearchBenchmarksMulti-objective RLHF alignment on safeRLHF (test)Multi-objective RLHF alignment on safeRLHF (test)Follow52Win RateMAVIS49.450.75253.3Aug 19, 2025Evaluation resultsMethodMethodLinksWin RateMAVISBase Model=Alpaca, Bas...Base Model=Alpaca, Baseline Compared Against=PARM2025.0852