ResearchBenchmarksDocstring Evaluation on DevEval 183 human-written docstringsFollow4.938ScoreHuman-0.166321.158842.4843.80916Apr 12, 2026Evaluation ResultsMethodMethodLinksScoreRank Correlation (rp)Rank Correlation (rs)Kendall's Tau (τ)Average Metric ValueHumanScore Range=1–5Score Range=1–52026.044.938————G-EvalScore Range=1–5Score Range=1–52026.043.0430.0130.0030.0030.006ReFEreeScore Range=0–1Score Range=0–12026.040.9380.3260.2980.2870.304ROUGE-LScore Range=0–1Score Range=0–12026.040.2080.0060.0470.0390.03BLEUScore Range=0–1Score Range=0–12026.040.030.0620.0610.0540.059