ResearchBenchmarksSkill Planning on 80 controlled natural-language augmentation requestsFollow92.5T1 ScoreSAGA53.563.62573.7583.875Jun 27, 2026Evaluation ResultsMethodMethodLinksT1 ScoreT3 ScoreDecision RateInversion RateObservation RateArgumentation RateRejection RateRecallFF ScoreSAGA2026.0692.597.592.509691.710087.581.2Rule2026.0670707057.991.881.27.963.750LLM2026.0661.266.261.210092.479.2051.230ReAct2026.0661.271.261.210096.979.605533.8Key2026.0655555510087.87604526.2