politeness / eval /calibration.json
bogdanraduta's picture
Upload eval/calibration.json with huggingface_hub
63c3e6f verified
Raw
History Blame Contribute Delete
12.7 kB
{
"threshold": 0.1,
"objective": "macro_f1",
"value_on_val": 0.9826,
"default_threshold": 0.5,
"default_value_on_val": 0.9817,
"improvement": 0.0009,
"chosen_on": "validation split, never test",
"curve": [
{
"threshold": 0.02,
"macro_f1": 0.9809,
"any_label_recall": 0.989,
"any_label_fpr": 0.0156
},
{
"threshold": 0.03,
"macro_f1": 0.9809,
"any_label_recall": 0.989,
"any_label_fpr": 0.0156
},
{
"threshold": 0.04,
"macro_f1": 0.9818,
"any_label_recall": 0.989,
"any_label_fpr": 0.0146
},
{
"threshold": 0.05,
"macro_f1": 0.9818,
"any_label_recall": 0.989,
"any_label_fpr": 0.0146
},
{
"threshold": 0.06,
"macro_f1": 0.9818,
"any_label_recall": 0.9872,
"any_label_fpr": 0.0135
},
{
"threshold": 0.07,
"macro_f1": 0.9818,
"any_label_recall": 0.9872,
"any_label_fpr": 0.0135
},
{
"threshold": 0.08,
"macro_f1": 0.9818,
"any_label_recall": 0.9872,
"any_label_fpr": 0.0135
},
{
"threshold": 0.09,
"macro_f1": 0.9818,
"any_label_recall": 0.9872,
"any_label_fpr": 0.0135
},
{
"threshold": 0.1,
"macro_f1": 0.9826,
"any_label_recall": 0.9872,
"any_label_fpr": 0.0125
},
{
"threshold": 0.11,
"macro_f1": 0.9826,
"any_label_recall": 0.9872,
"any_label_fpr": 0.0125
},
{
"threshold": 0.12,
"macro_f1": 0.9826,
"any_label_recall": 0.9872,
"any_label_fpr": 0.0125
},
{
"threshold": 0.13,
"macro_f1": 0.9826,
"any_label_recall": 0.9872,
"any_label_fpr": 0.0125
},
{
"threshold": 0.14,
"macro_f1": 0.9826,
"any_label_recall": 0.9872,
"any_label_fpr": 0.0125
},
{
"threshold": 0.15,
"macro_f1": 0.9826,
"any_label_recall": 0.9872,
"any_label_fpr": 0.0125
},
{
"threshold": 0.16,
"macro_f1": 0.9808,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0125
},
{
"threshold": 0.17,
"macro_f1": 0.9808,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0125
},
{
"threshold": 0.18,
"macro_f1": 0.9808,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0125
},
{
"threshold": 0.19,
"macro_f1": 0.9808,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0125
},
{
"threshold": 0.2,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.21,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.22,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.23,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.24,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.25,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.26,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.27,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.28,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.29,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.3,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.31,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.32,
"macro_f1": 0.9817,
"any_label_recall": 0.9835,
"any_label_fpr": 0.0114
},
{
"threshold": 0.33,
"macro_f1": 0.9808,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0114
},
{
"threshold": 0.34,
"macro_f1": 0.9808,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0114
},
{
"threshold": 0.35,
"macro_f1": 0.9808,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0114
},
{
"threshold": 0.36,
"macro_f1": 0.9808,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0114
},
{
"threshold": 0.37,
"macro_f1": 0.9808,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0114
},
{
"threshold": 0.38,
"macro_f1": 0.9808,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0114
},
{
"threshold": 0.39,
"macro_f1": 0.9808,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0114
},
{
"threshold": 0.4,
"macro_f1": 0.9808,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0114
},
{
"threshold": 0.41,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.42,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.43,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.44,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.45,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.46,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.47,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.48,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.49,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.5,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.51,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.52,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.53,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.54,
"macro_f1": 0.9817,
"any_label_recall": 0.9817,
"any_label_fpr": 0.0104
},
{
"threshold": 0.55,
"macro_f1": 0.9807,
"any_label_recall": 0.9798,
"any_label_fpr": 0.0104
},
{
"threshold": 0.56,
"macro_f1": 0.9807,
"any_label_recall": 0.9798,
"any_label_fpr": 0.0104
},
{
"threshold": 0.57,
"macro_f1": 0.9807,
"any_label_recall": 0.9798,
"any_label_fpr": 0.0104
},
{
"threshold": 0.58,
"macro_f1": 0.9807,
"any_label_recall": 0.9798,
"any_label_fpr": 0.0104
},
{
"threshold": 0.59,
"macro_f1": 0.9807,
"any_label_recall": 0.9798,
"any_label_fpr": 0.0104
},
{
"threshold": 0.6,
"macro_f1": 0.9807,
"any_label_recall": 0.9798,
"any_label_fpr": 0.0104
},
{
"threshold": 0.61,
"macro_f1": 0.9807,
"any_label_recall": 0.9798,
"any_label_fpr": 0.0104
},
{
"threshold": 0.62,
"macro_f1": 0.9807,
"any_label_recall": 0.9798,
"any_label_fpr": 0.0104
},
{
"threshold": 0.63,
"macro_f1": 0.9807,
"any_label_recall": 0.9798,
"any_label_fpr": 0.0104
},
{
"threshold": 0.64,
"macro_f1": 0.9807,
"any_label_recall": 0.9798,
"any_label_fpr": 0.0104
},
{
"threshold": 0.65,
"macro_f1": 0.9798,
"any_label_recall": 0.978,
"any_label_fpr": 0.0104
},
{
"threshold": 0.66,
"macro_f1": 0.9798,
"any_label_recall": 0.978,
"any_label_fpr": 0.0104
},
{
"threshold": 0.67,
"macro_f1": 0.9798,
"any_label_recall": 0.978,
"any_label_fpr": 0.0104
},
{
"threshold": 0.68,
"macro_f1": 0.9798,
"any_label_recall": 0.978,
"any_label_fpr": 0.0104
},
{
"threshold": 0.69,
"macro_f1": 0.9798,
"any_label_recall": 0.978,
"any_label_fpr": 0.0104
},
{
"threshold": 0.7,
"macro_f1": 0.9788,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0104
},
{
"threshold": 0.71,
"macro_f1": 0.9788,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0104
},
{
"threshold": 0.72,
"macro_f1": 0.9788,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0104
},
{
"threshold": 0.73,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.74,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.75,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.76,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.77,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.78,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.79,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.8,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.81,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.82,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.83,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.84,
"macro_f1": 0.9797,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0094
},
{
"threshold": 0.85,
"macro_f1": 0.9806,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0083
},
{
"threshold": 0.86,
"macro_f1": 0.9806,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0083
},
{
"threshold": 0.87,
"macro_f1": 0.9806,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0083
},
{
"threshold": 0.88,
"macro_f1": 0.9806,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0083
},
{
"threshold": 0.89,
"macro_f1": 0.9806,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0083
},
{
"threshold": 0.9,
"macro_f1": 0.9806,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0083
},
{
"threshold": 0.91,
"macro_f1": 0.9806,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0083
},
{
"threshold": 0.92,
"macro_f1": 0.9806,
"any_label_recall": 0.9761,
"any_label_fpr": 0.0083
},
{
"threshold": 0.93,
"macro_f1": 0.9797,
"any_label_recall": 0.9743,
"any_label_fpr": 0.0083
},
{
"threshold": 0.94,
"macro_f1": 0.9788,
"any_label_recall": 0.9725,
"any_label_fpr": 0.0083
},
{
"threshold": 0.95,
"macro_f1": 0.9788,
"any_label_recall": 0.9725,
"any_label_fpr": 0.0083
}
],
"notes": [],
"note": "**This threshold is not a tuned parameter, and the shipped policy default stays at 0.89.** Two seeds on the identical corpus read 0.10 and 0.36, a spread of 0.26, the same shape as nsfw's calibration on the same retrain campaign. Per-language quality is stable across the two seeds (mean spread 0.0015), the threshold picked from a flat curve is not. 0.89 is the value reviewed and shipped before this retrain and is unchanged by it."
}