Add calibrated thresholds (IdeaLens-1M calibration split)
Browse files- thresholds.json +83 -0
thresholds.json
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "rishanthrajendhran/ProseLens-ModernBERT-L",
|
| 3 |
+
"input": "document",
|
| 4 |
+
"flag_rule": "flag the document as AI when P(human) < cut",
|
| 5 |
+
"calibration": {
|
| 6 |
+
"data": "IdeaLens-1M, calibration split (human documents only)",
|
| 7 |
+
"scores_from": "the training run's stored predictions on the calibration split",
|
| 8 |
+
"n_humans": 80000,
|
| 9 |
+
"n_per_format": {
|
| 10 |
+
"Academic Writing": 10000,
|
| 11 |
+
"Creative Writing": 10000,
|
| 12 |
+
"Knowledge Article": 10000,
|
| 13 |
+
"News Article": 10000,
|
| 14 |
+
"Nonfiction Writing": 10000,
|
| 15 |
+
"Personal About Page": 10000,
|
| 16 |
+
"Personal Blog": 10000,
|
| 17 |
+
"User Reviews": 10000
|
| 18 |
+
}
|
| 19 |
+
},
|
| 20 |
+
"method": {
|
| 21 |
+
"global": "the target-FPR quantile of P(human) over all calibration humans",
|
| 22 |
+
"per_group": "the group's own quantile, shrunk toward the global cut with weight n / (n + 2500)",
|
| 23 |
+
"min_group_humans": 200,
|
| 24 |
+
"estimability": "a cut at FPR q is given only when q * n >= 25 calibration humans",
|
| 25 |
+
"missing_cut": "no cut for a group means none could be estimated; do not substitute the global cut"
|
| 26 |
+
},
|
| 27 |
+
"fpr_targets": [
|
| 28 |
+
0.001,
|
| 29 |
+
0.005,
|
| 30 |
+
0.01,
|
| 31 |
+
0.02,
|
| 32 |
+
0.05
|
| 33 |
+
],
|
| 34 |
+
"global": {
|
| 35 |
+
"0.001": 0.005766863141208887,
|
| 36 |
+
"0.005": 0.043511959370225665,
|
| 37 |
+
"0.01": 0.18790396720170977,
|
| 38 |
+
"0.02": 0.8414405298233032,
|
| 39 |
+
"0.05": 0.9982767522335052
|
| 40 |
+
},
|
| 41 |
+
"per_format": {
|
| 42 |
+
"0.005": {
|
| 43 |
+
"Academic Writing": 0.48511697150394323,
|
| 44 |
+
"Creative Writing": 0.7774797902218998,
|
| 45 |
+
"Knowledge Article": 0.0264915764220059,
|
| 46 |
+
"News Article": 0.042131727289408456,
|
| 47 |
+
"Nonfiction Writing": 0.016996091064065695,
|
| 48 |
+
"Personal About Page": 0.08180167488381268,
|
| 49 |
+
"Personal Blog": 0.42620853902027017,
|
| 50 |
+
"User Reviews": 0.24418430317565804
|
| 51 |
+
},
|
| 52 |
+
"0.01": {
|
| 53 |
+
"Academic Writing": 0.8101551976799966,
|
| 54 |
+
"Creative Writing": 0.83455596023798,
|
| 55 |
+
"Knowledge Article": 0.07479736334085464,
|
| 56 |
+
"News Article": 0.16216902357339863,
|
| 57 |
+
"Nonfiction Writing": 0.0544652368426323,
|
| 58 |
+
"Personal About Page": 0.36414291030168533,
|
| 59 |
+
"Personal Blog": 0.7883320108056069,
|
| 60 |
+
"User Reviews": 0.7249871950745583
|
| 61 |
+
},
|
| 62 |
+
"0.02": {
|
| 63 |
+
"Academic Writing": 0.9664362020492554,
|
| 64 |
+
"Creative Writing": 0.9677177896499634,
|
| 65 |
+
"Knowledge Article": 0.28712427091598514,
|
| 66 |
+
"News Article": 0.6937795658111573,
|
| 67 |
+
"Nonfiction Writing": 0.22745220160484314,
|
| 68 |
+
"Personal About Page": 0.9371422376632691,
|
| 69 |
+
"Personal Blog": 0.9637664785385132,
|
| 70 |
+
"User Reviews": 0.9604384603500367
|
| 71 |
+
},
|
| 72 |
+
"0.05": {
|
| 73 |
+
"Academic Writing": 0.9994967925548554,
|
| 74 |
+
"Creative Writing": 0.999563969373703,
|
| 75 |
+
"Knowledge Article": 0.9463144671916961,
|
| 76 |
+
"News Article": 0.9944453680515289,
|
| 77 |
+
"Nonfiction Writing": 0.8933786427974701,
|
| 78 |
+
"Personal About Page": 0.9992063510417939,
|
| 79 |
+
"Personal Blog": 0.9993082511425019,
|
| 80 |
+
"User Reviews": 0.999339975118637
|
| 81 |
+
}
|
| 82 |
+
}
|
| 83 |
+
}
|