Hugging Face's logo

rafmacalaba
/
gliner-datause-displacement

Token Classification
GLiNER
PyTorch
ner
data-use
tiered
Model card Files
xet
 Community
rafmacalaba commited on
Commit
ca74bcf
·
verified ·
1 Parent(s): 83aae15

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +11 -11
README.md CHANGED
@@ -41,16 +41,16 @@ matching, jaccard >= 0.5 — identical to prior data-use-mentions evals.
41
 
42
  | thr | tp | fp | fn | precision | recall | f0.5 | f1 | t3_leak | t3_leak% |
43
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
44
- | 0.10 | 14556 | 8850 | 2609 | 0.6219 | 0.8480 | 0.6569 | 0.7176 | 2400 | 27.1% |
45
- | 0.20 | 13394 | 5995 | 3771 | 0.6908 | 0.7803 | 0.7070 | 0.7328 | 1977 | 33.0% |
46
- | 0.30 | 12046 | 4436 | 5119 | 0.7309 | 0.7018 | 0.7249 | 0.7160 | 1686 | 38.0% |
47
- | 0.40 | 10514 | 3273 | 6651 | 0.7626 | 0.6125 | 0.7270 | 0.6794 | 1432 | 43.8% |
48
- | 0.50 | 8706 | 2304 | 8459 | 0.7907 | 0.5072 | 0.7112 | 0.6180 | 1163 | 50.5% |
49
- | 0.60 | 6762 | 1561 | 10403 | 0.8124 | 0.3939 | 0.6701 | 0.5306 | 918 | 58.8% |
50
- | 0.70 | 4605 | 950 | 12560 | 0.8290 | 0.2683 | 0.5846 | 0.4054 | 650 | 68.4% |
51
 
52
- **Best F0.5**: 0.7270 (thr=0.4)
53
- **Best F1**: 0.7328 (thr=0.2)
54
 
55
  Full per-doc predictions (raw scores, gold spans with tier decisions): `holdout_predictions.jsonl` on this repo.
56
 
@@ -58,5 +58,5 @@ Full per-doc predictions (raw scores, gold spans with tier decisions): `holdout_
58
 
59
  | corpus | examples | spans | thr | precision | recall | f0.5 | f1 |
60
  | --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
61
- | prwp | 7758 | 12359 | 0.40 | 0.7831 | 0.6601 | 0.7550 | 0.7164 |
62
- | fcv | 7574 | 9693 | 0.40 | 0.7158 | 0.5193 | 0.6655 | 0.6019 |
 
41
 
42
  | thr | tp | fp | fn | precision | recall | f0.5 | f1 | t3_leak | t3_leak% |
43
  | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
44
+ | 0.10 | 288 | 610 | 32 | 0.3207 | 0.9000 | 0.3681 | 0.4729 | 153 | 25.1% |
45
+ | 0.20 | 271 | 448 | 49 | 0.3769 | 0.8469 | 0.4240 | 0.5217 | 116 | 25.9% |
46
+ | 0.30 | 254 | 341 | 66 | 0.4269 | 0.7937 | 0.4704 | 0.5552 | 90 | 26.4% |
47
+ | 0.40 | 237 | 250 | 83 | 0.4867 | 0.7406 | 0.5225 | 0.5874 | 67 | 26.8% |
48
+ | 0.50 | 205 | 169 | 115 | 0.5481 | 0.6406 | 0.5644 | 0.5908 | 52 | 30.8% |
49
+ | 0.60 | 177 | 95 | 143 | 0.6507 | 0.5531 | 0.6286 | 0.5980 | 36 | 37.9% |
50
+ | 0.70 | 128 | 44 | 192 | 0.7442 | 0.4000 | 0.6349 | 0.5203 | 19 | 43.2% |
51
 
52
+ **Best F0.5**: 0.6349 (thr=0.7)
53
+ **Best F1**: 0.5980 (thr=0.6)
54
 
55
  Full per-doc predictions (raw scores, gold spans with tier decisions): `holdout_predictions.jsonl` on this repo.
56
 
 
58
 
59
  | corpus | examples | spans | thr | precision | recall | f0.5 | f1 |
60
  | --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
61
+ | prwp | 0 | 0 | 0.10 | 1.0000 | 1.0000 | 1.0000 | 1.0000 |
62
+ | fcv | 0 | 0 | 0.10 | 1.0000 | 1.0000 | 1.0000 | 1.0000 |