ResearchBenchmarksVerified Code Gen. on BigCodeBench (test)Follow38.97Pass RateTEP27.332430.353733.37536.3963Jan 28, 2026Evaluation ResultsMethodMethodLinksPass RateTEPiterations (free phase...iterations (free phase)=20, iterations (nudged phase)=402026.0138.97TextGradsummarization=falsesummarization=false2026.0135.71TextGradsummarization=truesummarization=true2026.0135.12CoT2026.0134.15DSPy2026.0133.81HBC2026.0127.78