ResearchBenchmarksJob scheduling with temporary disruptions on τ2-bench+ / REALM-bench+ P12 v1 (test)Follow3Success CountLG(PE)-0.120.691.52.31May 5, 2026Evaluation ResultsMethodMethodLinksSuccess CountTokens Consumed (k)Time Taken (s)Completion CountLG(PE)Framework=LG(PE)Framework=LG(PE)2026.05317,00059—RAC_MFramework=RAC_MFramework=RAC_M2026.05310,00021—RACFramework=RACFramework=RAC2026.0539,00018—SagaLLMFramework=SagaLLMFramework=SagaLLM2026.05171,0005321LGFramework=LGFramework=LG2026.0501,00011—