ResearchBenchmarksTool Use on BFCL (Live/Non-live/Overall)Follow69.9Live Success RateCHORD-ϕ58.77261.66164.5567.439Aug 15, 2025Evaluation ResultsMethodMethodLinksLive Success RateNon-live Success RateOverall Success RateCHORD-ϕVariant=ϕVariant=ϕ2025.0869.990.278.5CHORD-µVariant=µVariant=µ2025.0869.488.677.6GRPOMode=Pure RLMode=Pure RL2025.0868.588.877.1SFT-best + RLType=SFT and RLType=SFT and RL2025.0867.487.976.1LUFFY2025.0867.28876.1SRFT2025.0864.685.873.6SFT-bestType=SFTType=SFT2025.0859.284.269.8