ResearchBenchmarksReasoning on Formula (test)Follow86.5AccuracySAGE66.2271.48576.7582.015Jun 17, 2026Evaluation ResultsMethodMethodLinksAccuracySAGE2026.0686.5SPO-RSruns=3runs=32026.0685.5SPO-GAruns=3runs=32026.0685.5ACE offline2026.0685GEPA2026.0671.5MIPROv22026.0669.5DC-CU2026.0669.5Base LLM2026.0667.5ICL2026.0667