Repository Public Research preview
1\begin{tabular}{l r c c c c c}2\toprule3System & $n$ & chrF++ & edit similarity & ROUGE-L & number recall & placeholders exact \\4\midrule5B0 identity & 15649 & 68.91 [68.35, 69.48] & 0.682 [0.677, 0.687] & 0.689 [0.683, 0.695] & 1.000 [1.000, 1.000] & 1.000 [1.000, 1.000] \\6B1 paraphrase & 15649 & 66.23 [65.68, 66.78] & 0.569 [0.564, 0.575] & 0.605 [0.599, 0.612] & 1.000 [1.000, 1.000] & 1.000 [1.000, 1.000] \\7B2 flan-t5-large & 15649 & 70.71 [70.19, 71.24] & 0.700 [0.695, 0.705] & 0.713 [0.708, 0.719] & 0.987 [0.983, 0.990] & 0.992 [0.989, 0.995] \\8B2 flan-t5-xl & 15649 & 71.30 [70.79, 71.81] & 0.704 [0.699, 0.709] & 0.718 [0.713, 0.724] & 0.986 [0.982, 0.989] & 0.992 [0.988, 0.994] \\9B2 t5-small & 15649 & 68.83 [68.31, 69.34] & 0.684 [0.679, 0.689] & 0.693 [0.688, 0.699] & 0.986 [0.983, 0.990] & 0.733 [0.706, 0.760] \\10LLM qwen2.5-1.5b-s2 & 15649 & 70.89 [70.34, 71.43] & 0.698 [0.693, 0.703] & 0.710 [0.705, 0.716] & 0.998 [0.997, 0.999] & 0.998 [0.997, 0.999] \\11LLM qwen2.5-1.5b & 15649 & 70.88 [70.33, 71.42] & 0.698 [0.693, 0.702] & 0.710 [0.705, 0.715] & 0.998 [0.997, 0.999] & 0.998 [0.997, 0.999] \\12LLM qwen2.5-3b & 15649 & 71.18 [70.64, 71.73] & 0.701 [0.696, 0.706] & 0.713 [0.708, 0.719] & 0.998 [0.997, 0.999] & 0.999 [0.998, 1.000] \\13edit-tiny & 15649 & 66.71 [66.12, 67.30] & 0.675 [0.670, 0.680] & 0.706 [0.701, 0.712] & 1.000 [1.000, 1.000] & 0.999 [0.999, 1.000] \\14m1-base-lambda-s2 & 15649 & 69.54 [69.01, 70.07] & 0.685 [0.681, 0.690] & 0.695 [0.690, 0.701] & 0.991 [0.989, 0.993] & 0.992 [0.989, 0.994] \\15m1-base-lambda & 15649 & 69.60 [69.06, 70.15] & 0.687 [0.682, 0.692] & 0.697 [0.691, 0.702] & 0.992 [0.991, 0.994] & 0.993 [0.991, 0.995] \\16m1-base-lr2e5 & 15649 & 69.46 [68.93, 69.99] & 0.685 [0.680, 0.690] & 0.694 [0.689, 0.700] & 0.991 [0.989, 0.993] & 0.992 [0.989, 0.995] \\17m1-base-mps & 15649 & 68.65 [68.13, 69.18] & 0.677 [0.672, 0.682] & 0.686 [0.680, 0.691] & 0.988 [0.985, 0.989] & 0.990 [0.988, 0.993] \\18m1-large-lambda-stage3 & 15649 & 67.66 [67.14, 68.19] & 0.666 [0.661, 0.671] & 0.674 [0.669, 0.680] & 0.983 [0.980, 0.986] & 0.984 [0.980, 0.988] \\19m1-large-lambda & 15649 & 68.20 [67.66, 68.76] & 0.671 [0.666, 0.676] & 0.680 [0.674, 0.685] & 0.988 [0.986, 0.990] & 0.987 [0.983, 0.991] \\20m1-large-lr15e6 & 15649 & 68.12 [67.60, 68.66] & 0.672 [0.667, 0.677] & 0.680 [0.675, 0.686] & 0.988 [0.986, 0.990] & 0.989 [0.986, 0.992] \\21m1-small-v2 & 15649 & 68.82 [68.31, 69.34] & 0.678 [0.674, 0.683] & 0.686 [0.681, 0.692] & 0.988 [0.986, 0.990] & 0.989 [0.986, 0.992] \\22m1-small & 15649 & 69.02 [68.50, 69.53] & 0.679 [0.675, 0.684] & 0.688 [0.683, 0.694] & 0.990 [0.988, 0.992] & 0.991 [0.988, 0.993] \\23m1-tiny-s2 & 15649 & 69.30 [68.76, 69.84] & 0.685 [0.680, 0.690] & 0.694 [0.689, 0.699] & 0.991 [0.989, 0.992] & 0.992 [0.990, 0.995] \\24m1-tiny-s3 & 15649 & 69.35 [68.82, 69.88] & 0.684 [0.680, 0.689] & 0.693 [0.688, 0.698] & 0.991 [0.990, 0.993] & 0.992 [0.990, 0.994] \\25m1-tiny & 15649 & 69.31 [68.78, 69.85] & 0.684 [0.679, 0.689] & 0.693 [0.687, 0.698] & 0.991 [0.989, 0.992] & 0.992 [0.990, 0.995] \\26m2-tiny-s2 & 15649 & 69.37 [68.85, 69.91] & 0.686 [0.681, 0.691] & 0.694 [0.689, 0.700] & 0.991 [0.989, 0.993] & 0.992 [0.990, 0.995] \\27m2-tiny & 15649 & 69.17 [68.64, 69.72] & 0.683 [0.678, 0.688] & 0.692 [0.686, 0.697] & 0.989 [0.986, 0.991] & 0.993 [0.991, 0.995] \\28m3light-tiny & 15649 & 69.00 [68.46, 69.54] & 0.682 [0.677, 0.687] & 0.690 [0.685, 0.696] & 0.989 [0.987, 0.990] & 0.993 [0.990, 0.995] \\29m3tokenonly-tiny-s2 & 15649 & 69.23 [68.69, 69.76] & 0.684 [0.680, 0.689] & 0.693 [0.688, 0.699] & 0.991 [0.989, 0.993] & 0.993 [0.990, 0.995] \\30m3tokenonly-tiny-s3 & 15649 & 69.44 [68.90, 69.98] & 0.686 [0.682, 0.691] & 0.695 [0.690, 0.701] & 0.991 [0.990, 0.993] & 0.993 [0.991, 0.995] \\31m3tokenonly-tiny & 15649 & 69.31 [68.77, 69.86] & 0.686 [0.681, 0.690] & 0.694 [0.689, 0.700] & 0.991 [0.989, 0.992] & 0.994 [0.991, 0.996] \\32\bottomrule33\end{tabular}34